Oral evidence - The internet: to regulate or not to regulate?

Hugh Milward, Katie O’Donovan and Rebecca Stimson.

Q174 The Chairman: Good afternoon. May I welcome our witnesses to this session of the House of Lords Communications Committee on our inquiry into internet regulation? Our witnesses are Rebecca Stimson, Katie O’Donovan and Hugh Milward, and I will ask them to introduce themselves in a moment. I would remind you that we are recording today’s session. It will be broadcast online and a transcript will be prepared. There is the possibility of a Division this afternoon and, if that occurs, we will briefly suspend the meeting and resume after 10 to 15 minutes.

May I ask our witnesses from the tech giants briefly to introduce themselves and tell us a bit about the perspectives of their organisations? In doing that, perhaps they would answer a couple of initial questions. What are the advantages and disadvantages of the current regulatory framework for the internet as it affects your businesses and society? Also, may I ask our witnesses for their reactions to the Chancellor’s announcement yesterday of a digital services tax, which is clearly targeted at the businesses that our witnesses represent? Perhaps we can start with Rebecca Stimson.

Rebecca Stimson: I am head of public policy for Facebook in the UK. I have been with Facebook for coming up to a year. I was a civil servant in the UK for 20 years prior to that, most recently in the Ministry of Justice. I do not have an opening statement other than to say I am very happy to be here as part of this very important inquiry.

Your first question was about the advantages and disadvantages and I think the advantages of the current framework are pretty clear. The UK has quite a thriving digital economy. It attracts an enormous amount of tech investment. It has an admired regulatory framework that applies to all our companies in various different ways. The evidence speaks for itself, if we think about consumers and the amount of choice they have, and the different platforms that they are able to use and engage with. What we see suggests that the framework is working extremely well.

I am sure we will come on to some of the disadvantages in the questions around how that framework can keep pace with change. Having done some work on GDPR in my old life as a civil servant, I know how complicated it is to write something that is future-proofed.

In answer to your tax question from yesterday, we are in the process of looking at exactly what the Chancellor proposed, and we need to see it in detail, but there is a consultation document expected very soon, as I understand it, and we will be fully engaging with that process when we see it.

Katie O'Donovan: I am UK public policy manager at Google with responsibility for a number of our policy areas in the UK. I also sit on the board of the Internet Watch Foundation. I mention that as it may come up in the course of the conversation.

Thank you very much for inviting us to give evidence today. We have been following your inquiry with interest. We recognise that this is a time where people are thinking more and more about how they use technology in their daily lives and, indeed, whether regulation has kept pace and whether there are areas to explore within those conversations. While Google is institutionally very young—we recently turned 20—as a technology company, we are one of the oldest. That gives us a perspective from which to see how our products have evolved over time, how the use from users has changed over time, and whether we need to change the way that we exist and operate.

In terms of the benefits and shortcomings of the current situation, there are inherent benefits in the way we are able to access the internet online in the UK, and sometimes we skip too quickly over them. We have the ability to access information from across the world in almost real time. A school child in India and a professor in Oxford now have the same ability to find out crucial information, to stay in touch with friends and family from around the world and to start businesses and exporting, in a way that was impossible to think of 10 or 15 years ago. It is well worth contemplating that and preserving it.

As I reflected earlier, the way that we all use the internet in our daily lives, the way that young people and companies use the internet has changed quite dramatically. We have changed and evolved as a company. We have been able to address some of the issues in the way we work without waiting for regulation. If you think about the way that we deal with very serious issues such as child sex abuse imagery, we have been able to work in partnership with institutions such as the Internet Watch Foundation. We also work very closely with the Government and other bodies when there are issues that require specific regulation. It is an inquiry that we are very happy to participate in and we have learned a lot from the evidence of your other sessions, too.

On the question of the Budget yesterday, we also saw the Chancellor’s proposal and are waiting for the consultation document. We very clearly understand the importance of the tax issue and the policy discussions and scrutiny that have come under that. The Chancellor referred to his proposals as part of setting a timeline for international action and we have always supported that. For a tech company such as ours, and indeed for many other companies that operate across borders in different countries, a multilateral international solution would be really meaningful and of long-term significance. We continue to support an international resolution to these issues.

The Chairman: Hugh Milward.

Hugh Milward: I run corporate, external and legal affairs for Microsoft, but I am not a lawyer. I want to make that really clear. Microsoft seems to be one of the elder statesmen of technology companies. We are nearly 45 years old and we have seen a few battles over the years. Some of these issues have arisen before and we are very keen to participate in your Committee and try to offer what we can from the experience we have gleaned over the years.

The opportunities that technology affords society are tremendous and very significant, especially with the advent of AI, something that we are developing at pace. The biggest worry that we have is around the trust that society has in technology. We believe firmly that if society does not trust the technology, it will not the use it, and will not benefit from the opportunities that technology provides. It is incredibly important that we get this right and ensure there is a high level of trust across society, and that we come together as the technology industry in committees such as this, and with government and civil society, to navigate our way through some of these very complex issues. These issues are developing at pace and in real time and we are trying to find ways in which to solve a series of issues that society is concerned about.

At the heart of it is how we make sure that society can trust the technology that is going to benefit them so much. We do not feel that there is a Wild West of unregulated space at the moment. In fact, there are a range of regulations in place which help to provide that level of trust. If you look at laws governing connectivity, intellectual property, copyright, net neutrality, data protection, privacy, advertising standards, et cetera, these are all regulations and laws that already affect technology companies. Generally, if a law applies offline it applies online as well. It is not really the Wild West that it is sometimes painted. There are some very specific examples of quite new regulations or voluntary measures that are working extremely well, which we can go into in due course.

In terms of a digital tax, again we are digesting what it was that the Chancellor announced yesterday. It certainly looks to be interesting. What remains to be seen is how this would dovetail with what the OECD is driving at. Probably the most important thing is how this influences what the OECD is thinking and how the OECD influences what the Chancellor decides to do. We will respond to the consultation, as the others have said.

Q175 Lord Allen of Kensington: I would like to stick with tax. I completely understand you saying that you need to understand the detail, but my question is more philosophical, in that to be trusted, as you said, Hugh, and to be a good corporate citizen, frankly, tax is a big issue, whether in the pub or in Parliament, and people do not understand why you are taking so much revenue from the UK and you are paying so little tax because of clever tax schemes. You might say that you are paying what you are asked to pay, but that does not feel like you acting as a trusted good corporate citizen. I would like your views not on the detail of the tax but that specific point, because it is a massive issue. A number of people who have given evidence have raised that. The second point, and related to that, is whether 2% is equitable versus what non-digital companies pay; does that feel equitable? The third thing is a number of people have said, whether it is a tax or a levy, something could be used to help fund the regulation of the internet. I should like your views on those three points.

Rebecca Stimson: We all recognise that tax is quite a sensitive issue and, as you would expect me to, I would say that we pay all the taxes in the UK that we are required to pay. In recognition of your point that people look at the turnovers of these companies and have questions about the tax regime, I am sure you are aware that Facebook made a change in 2016 to move more to a local-seller model to increase the amount of tax that we pay in the UK.

Lord Allen of Kensington: What percentage of your turnover is paid in tax?

Rebecca Stimson: I would have to check that figure. I do not know that figure off the top of my head, I am afraid. To your point about a levy, I can again understand why people might look at companies such as Facebook and say, “We should get them to pay a levy”. That is not an idea that we would automatically be against, but, as you have heard from previous witnesses, and certainly from the way the Government are approaching this whole area, we need to think about what you are trying to do and the harms you are trying to tackle. Recently, the head of the NHS, for example, called for a levy to address the impact on mental health of social media. We would need to begin by looking at the evidence, scoping that problem and defining the harms, and work through to whether it is clear that a levy on social media companies would be the most effective way forward.

Katie O'Donovan: You are absolutely right that this is an issue that consumes people whether they are in a pub or in Parliament and they often want to discuss it. We also pay all the tax that we are due to pay in the UK and that has increased over the years. The question is about how we pay the tax as a proportion globally. We are a US-founded company which is headquartered in the US, and so our global tax rate over the last decade has been 26%, which is comparable to UK corporation tax, but the proportion we pay in the US as our home country is 80%. We think it is important to have an international resolution to this tax issue so that the issue is not solved in one country but has knock-on consequences in other countries.

Hugh Milward: Likewise, we pay all the tax that we owe in the UK, as we should. Before we start looking at a levy to fund the regulation of the internet, we would probably want to go back to thinking about what kinds of interventions are required to get the desired outcomes through regulation of the internet, and look at how we design that to solve the problems that we are trying to solve; and then look at whether a regulator is the right approach and how we would fund a regulator, rather than doing it the other way round.

Lord Allen of Kensington: I would ask the same question of Katie about funding regulation. I would also ask each organisation what percentage of turnover is paid in tax in the UK. You might need to come back to us.

Hugh Milward: I will need to come back to you on that.

Katie O'Donovan: The tax we pay on our profits globally is 26%.

Lord Allen of Kensington: I am talking about tax in the UK as a percentage of turnover.

Katie O'Donovan: We can come back to you on that.

The Chairman: We will write to all three of you and ask you to tell us the amount of tax that you pay in the UK as a percentage of your turnover in the UK.

Katie O'Donovan: That is not generally how tax is calculated.

Lord Allen of Kensington: I understand that, but I am trying to understand how equitable it is versus other offline companies. Katie, you were going to come back on funding regulation.

Katie O'Donovan: You asked about a levy. Rebecca mentioned one example but there are lots of different areas where, in recent times, a levy has been suggested to pay for whether it is a regulator or a particular part of the service. Indeed, the Government’s Green Paper on the internet safety strategy suggested a levy around educational services for online safety for children. We have discussed this with the Government. In that particular instance, we invested millions of pounds in our own education programme. We have a programme called Internet Legends, and one for teenagers called Be Internet Citizens, which reach hundreds of thousands of young people each year. To build on Hugh’s point, we need to have a very specific point of reference for what a regulator would do or what the levy would be required to do, and if that is the most effective way to do it. It is certainly a conversation that we are very happy to continue with government.

The Chairman: Baroness Quin and then we will move on to market concentration.

Q176 Baroness Quin: Hugh mentioned trust and that is an important theme that runs through a lot of the work that we have been doing here. Does each of your companies operate under a set of guiding principles for how you deal with customers and how such principles can strike a balance between your desire to get as much information from customers as possible and, at the same time, respect the desire of customers not to want to have too much information about them being divulged or accumulated? If you have such guiding principles, are they made public or are they discussed between the main companies which are operating? How do you interact with, say, other people who have given evidence, such as Which? or Doteveryone, in terms of addressing their concerns? I know those are wide-ranging questions, but could you at least make a start on the guiding principles and what this means for users and people who are concerned about users?

Rebecca Stimson: We want Facebook to be a safe and enjoyable place where all the people who sign up to use it get a positive experience to connect with the people who mean something to them—their friends and family. There are a whole host of guiding principles that we operate under depending on what aspect of the business you are talking about. It can be anything from ensuring that our policies are very clear and accessible to people, so things such as data privacy and what we do with your information, to educational material about how to be safe online. We have a number of different ways that we engage with parents, teachers and younger people to try to drive up digital literacy and awareness. Sometimes that is the quite practical basics of understanding how to do certain things on our platform. We try to design our tools so they are intuitive and easy for people to use. You referenced people’s information, and the way it is framed and reflected in regulation is that people’s data is theirs. They can control it, move it around, set their privacy settings and manage their data on our platform with a whole range of tools that we provide. As I say, we embed the values of our company into the products that we make and we try, in a whole host of ways, to reflect that for the consumers who use them.

Katie O'Donovan: Google has a singular company mission, which is to make the world’s information universally and usefully accessible. That guides everything that we do. We also have a very practical principle in that we put the user first in all the decisions that we make. To give you an idea of how that comes to life, when Google started as a search engine, the creators wanted to get people off the search engine as quickly as possible. One reason that Google was successful is that you could search for your answer and you knew with confidence that you could click on the blue link and it would take you to the site that you were looking for. Back when Google first started, other search engines created incentives for you to stay on the search page site. You sometimes had quizzes or crosswords or Sudoku alongside the search engine, because part of their business model was to keep you on their search engine site for as long as possible. We know we are doing our best as a search engine, and it is inherent to the users’ experience that they know they can go to Google and within a fraction of a second they can click on the sites they want. That is not necessarily commercially viable. On the vast majority of searches we do not have advertising alongside because they are searches that people do not want to advertise against. On the searches where we believe there is a commercial attraction to advertise, or where people want to advertise, they can, and, of course, if somebody clicks on their link we receive advertising revenue from that.

By putting the customer and user first, we were able to build a fast and efficient search engine which was not monetised in every single search, but which did such a good job customers and users kept coming back to it. You mentioned Doteveryone and Which?. Which?, particularly, has a long history of understanding consumer rights and behaviour in the UK and Doteveryone has done some very interesting work in this area. To have a bit more granularity and understand a little more about how a search works, we published what we call our rater guidelines, which is a 160-page or so document, which is freely available on the internet to everyone. It is a set of guidelines we give to people who do quality control work on search engines to check that our algorithms are doing what is best for the user, best for the searcher, and makes the right decisions to return the right results. Those guidelines are available for anyone to look at and for a research organisation to fully scrutinise if it wants to.

Hugh Milward: We have a set of principles at the heart of our mission as well, which is about empowering everyone on the planet to achieve more.

Baroness Quin: Are those principles made public?

Hugh Milward: Yes. For the most part, the services that Microsoft offers are services that people pay for—Word, Excel, Powerpoint and a variety of others—and if you are looking at some of the more leading-edge services, in AI for example, again we are building a series of building blocks that customers use for their own purposes. Most enterprise customers use them for their own purposes and they pay us for it. We offer it in that way. In that sense, they see a direct value in the information and service they get, otherwise they would not buy it. That monetary exchange is very transparent.

The principles that we believe in around data, very similar to what my colleagues on the panel have said, are about it being the users’ data and that they should have control over that data, determine what happens to it, where it goes, where it is located and how it is treated. That is a fundamental principle of the way in which we design and offer services for our customers.

Baroness Quin: I should have made clear a family interest in that my stepson works for Google. I take the points that you made in response. We seem to have come across a general perception that people are nervous about what data is held about them. It does not seem to them as transparent and as open as you have suggested. As a user myself, if I go online and am asked about cookies, for example, I tend to say, “Yes, that’s all right”, without thinking, because I want to get the information quickly, and I wonder afterwards whether I should have done that or not. Despite the good intentions of the principle, is there still a gap between what people know about the system and how to access their own data and so on; and despite the procedures that you put in place, is there a problem of users not being familiar with the ways that they can protect themselves?

Katie O'Donovan: Which? has done some research into this to show a disconnect between the information that is available around their data and how they understand it. For example, if you have a Google account, you can go on to something called My Account and it will show exactly what data we have and how we use it. You can choose how we use it and if you want us to understand where you are searching from, you can share that. If you do not, you can turn that off. Globally, we have had 2 million visits to that, which is an extraordinarily high number, and a positive measure for us to say that that is working.

We also see a very high number of people engaging with that. There absolutely is more that can be done to make us conscientious users of technology. One reason we have invested so heavily in primary and teenage education is to help people understand online literacy and begin to think more consciously and critically about how they share information, and in which cases that is beneficial and in which cases it is not.

Hugh Milward: An incredibly important principle lies behind this, which is that if you have an accident and an ambulance comes to pick you up and take you to a hospital that you have never been to before, you really hope that the medical practitioners who are going to be caring for you have access to your medical records and to other bits of information that you have previously given to the NHS, so that they can treat you in the best possible way. If the consumer, if society does not have—and I go back to my point about trust—a level of trust that allows the NHS to have that data to be able to use it for the benefit of patients, something has gone wrong.

It is incredibly important that we ensure that we separate out the different concerns people have about different types of data and the way that the data is used. It is a lot more nuanced than a simple, “Do we or do we not trust other entities of whatever kind with our data?”, because there are several use cases where we can very clearly see how consumers would absolutely trust different entities with their data. I have nothing against the NHS at all, but the way the NHS is currently looking after data raises a lot of questions because its storage of data includes manila envelopes on trolleys in corridors. There is a lot further we can go and a lot of trust that we need to continue to build with the general public about the use of data that is not about stoking fears around data use.

The Chairman: Rebecca, do you wish to add anything?

Rebecca Stimson: No, I would reflect the same kinds of comments.

The Chairman: All three witnesses have referred to published guidelines or principles that they have. The Committee would welcome it if you could send what you have published and an indication of where you publish it and in what way it is available, and in what way you measure your conduct and performance against it. Baroness Chisholm.

Q177 Baroness Chisholm of Owlpen: Sir Tim Berners-Lee has expressed concerns that the world wide web has “evolved into an engine of inequity and division, swayed by powerful forces who use it for their own agendas”. Are there any risks for consumers and citizens associated with the concentration of digital markets within the hands of a few large tech companies? If so, how might such risks be mitigated?

Katie O'Donovan: That reflection is really important and one that we should—and do—consider very carefully. The way that we all use the internet now is very different from how we used it 20 years ago, and it continues to change. I do not recognise the characterisation of concentration of the digital market in the way that it is commonly portrayed. We have a very clear mission as a company and we operate with great transparency on this. The way that we add value to people in the UK can be measured at an economic value of around £50 billion. That economic value, the low barriers to entry and the innovation that the internet is able to provide are worth reflecting on. There is also a highly competitive market online because of those low barriers to entry. I talked a little about how searches evolved from 20 years ago, when we first started, and even today, around a quarter of searches are brand new; they are for information that people were not looking for yesterday or are looking for in a distinctly different way.

That provides us with an enormous challenge. Even if everything else had stayed still, we need innovation to take us from where our search engine was 20 years ago, when the amount of information on the web was akin to a big city or university library, to the exponential growth of information today, where users not only require search results as quickly as they did 20 years but they want them from their mobile phone or they want to use voice or they want something else. That change in technology and in consumers’ expectations delivers innovation in our markets. In the markets that Google operates in in the UK, when people go online to buy something, 50% of those journeys start on Amazon. We are competing with travel organisations for flight information and other video platforms which are launching, and there is a real vibrancy to that.

Over the 20 years that Google has existed, companies have been at the top of their game and then fallen away. The prominence of certain companies at a certain time does not reflect a lack of innovation in that space or a certainty over what will happen going forward. We thrive on innovation and feel we are operating in a very competitive environment.

Rebecca Stimson: I would reflect that very much. To use Katie’s words of a thriving digital economy in the UK, while there are bigger and smaller players, if you think about it from a consumer perspective, and I am sure I read recently that on average people have 80 apps on their phone—I do not but apparently people do—it has never been easier to start up these kinds of businesses. In the time that Facebook has existed, other multimillion-dollar international companies have either grown alongside us, such as Twitter, or grown with us, such as Spotify. There are tens of millions of UK businesses that operate successfully through our platform. Part of the regulatory framework now—to reflect the previous question—concerns the fact that the data that flows around to enable that is the users’ data, and it is theirs to move around. If I think about my own phone, I give my data to numerous apps that I use on that phone. It is not the preserve of any one company, irrespective of its size in the market. I would echo what Katie has said, that the premise of the question is certainly not our experience of being in a very competitive market at the moment.

Hugh Milward: The only thing I would add is that the increasing trajectory of digitisation in the economy means that pretty much every company is a tech company, or should be a tech company within a few years. In that kind of market you have to look again at what you mean by competition. What does dominance look like? If you look at operating systems, for example, around 10 years ago we went from a market penetration of 90%-odd down to around 14% of the install base, and that is because of the arrival of a whole variety of different competitor operating systems which fundamentally changed the market. That was over a nine-month period. You see these fundamental shifts in the way that we think about what competition looks like, and with the level of digitisation across all major companies in the UK economy now, we think pretty much every company will be a tech company, and that will mean big changes and a lot of competition in the marketplace.

Baroness Bonham-Carter of Yarnbury: I take the point that there are bigger and smaller players, but there are the really huge players that you all represent. What are the implications of such a tiny number of companies acting as gatekeepers to the internet? We heard from the Information Commissioner that she was concerned at the “pervasiveness of big data analytics and micro targeting. These concerns are magnified by mergers and acquisitions where personal data is the primary asset”.

Rebecca Stimson: Partly I would go back to something I said previously, which is that I would look at this from the consumers’ perspective. We have a very robust and well-established competition law framework and regulatory framework in this country and, if you think about how they approach these things, they look at the conduct of the companies and whether they are abusing their market position.

Baroness Bonham-Carter of Yarnbury: Who does?

Rebecca Stimson: The competition authorities.

Baroness Bonham-Carter of Yarnbury: Sorry, I thought you were talking about the consumer.

Rebecca Stimson: Not yet. They look at whether we are abusing our market position, whether we are a barrier to entry, whether we are upholding proper standards of ethics of safety around users’ data, whether consumers have lots of choice. As far as I am aware, the relevant authorities in the UK are satisfied that that is what this digital economy looks like in the UK at the moment. As I have said, people worry about data being concentrated in particular companies, but, to reflect Hugh’s point, the way that the markets are evolving means that data is shared by consumers with all kinds of different companies, from a supermarket online, to a social media app, to Uber, to something else, and it is not the preserve of those companies to hold it or fence it away from anyone; you can move it around as much as you like. I would approach the question in that sense, thinking about competition regulation as it stands at the moment, the way in which it comes at it from a consumer perspective and the very positive picture there is in the UK at the moment.

Baroness Bonham-Carter of Yarnbury: So you do not accept the Information Commissioner’s concerns.

Rebecca Stimson: I think the Information Commissioner was talking about micro targeting; is that correct?

Baroness Bonham-Carter of Yarnbury: Yes, which is quite an important element in this.

Rebecca Stimson: I understand that as being a slightly different question from competition—and please tell me if I have misunderstood that. That is about how messages, campaigns and advertising target people online.

Baroness Bonham-Carter of Yarnbury: That was my question.

Rebecca Stimson: Apologies if I misunderstood your question. People are very interested in the issue of how advertising is targeted online. At Facebook we have been quite transparent about how our algorithm operates to inform what people see in their newsfeed. As I am sure you saw, this is a particular issue when it comes to political advertising, and we made a change, I believe it was last week, that, going forward, all political adverts have to be labelled as such and it has to be clear who is paying for them. We have done that in advance of a consultation the Cabinet Office is currently doing on the same kinds of reforms to electoral law, particularly in the space of advertising. We have already made a change in advance of that consultation.

Katie O'Donovan: On concentration of data, to reflect on some of those points, data is not a limited asset. It is not like a physical property. To tackle a phrase that is often used, it is not like oil that only one person can use and own. People control it themselves, and that has been strengthened by the GDPR, and certainly, as a company, we never sell data to any other company. We empower our users to manage their own data. Another thing that is relevant to how we operate as a company is the technology and methodology that we use on data that adds the value and enables us to offer services that people keep asking to use. As Hugh mentioned earlier, the use of artificial intelligence has become increasingly important in our work and, as such, over the course of the last couple of years we have re-trained all our engineers so they can be artificial intelligence-based in their approach to topics.

Baroness Bonham-Carter of Yarnbury: That does not fill me with great confidence, I have to say.

Katie O'Donovan: AI suffers from the fact that it can be portrayed in a way that is geeky and fantastical—the stuff of futures, but not necessarily the futures that all of us would like. In practical terms, however, it can really help people. We use it in maps to help us understand if it is quicker to go from here by walking up to Tottenham Court Road or to get on the Northern line and leave at Charing Cross.

Baroness Bonham-Carter of Yarnbury: Katie, does it not send people deeper and deeper into where they are wanting to be sent rather than expanding?

Katie O'Donovan: Not at all. The way we use artificial intelligence is to enable users to perform the tasks they want to do quicker and more effectively. We have developed our own AI principles to ensure that we use them ethically, that we have transparency about them and that we use them for social good. It is important to say that AI could be misrepresented as being able to be used for sinister ends, but it is a practical technology that can be used positively.

My point was that as we have increased our capability in artificial intelligence as a company, we have made that publicly available and open source, so that engineers from other companies, whether they are competitors or not, or a computer studies student in their bedroom, can access TensorFlow, which is our open source artificial intelligence, and build their own programs from that. We are not keeping that technology to ourselves. We are broadening the whole ecosystem with that. It is an incredibly popular service that we offer on the open web.

Hugh Milward: Along the same lines, we are designing for others to use rather than for us to use. It is a slightly different model, I guess. There are different types of data and, as has previously been mentioned, there is very little data that is unique and proprietary and cannot be replicated easily anywhere else. For the vast majority of data you can create observed or inferred datasets quite easily, and that results in very low barriers to entry for new market entrants. That is one of the tests that competition authorities look at. In those situations where there are unique datasets, and where there is no substitute for them, it is right that the competition authorities look at that and test whether it is a barrier to competition.

The Chairman: May I ask a question on competition more generally? Which? says that many users of your services regard you as utility services that they cannot do without. Do you therefore understand why it is argued that you should be regulated as utilities?

Hugh Milward: Users quite like Word, Outlook and various things, but there are free versions of everything we offer, and when they decide to pay for another year’s subscription, they do so in the knowledge that there are a lot of free alternatives. I think the market is working well there. If there is a dominance of Word, Outlook and other things, it is a dominance because people choose it again and again.

The Chairman: Do you see yourselves as a utility provider?

Katie O'Donovan: No, I do not think we do, because there is a real choice. Every time people go online, there are rival search engines that have grown phenomenally over recent years, operating in a different way from us. As I mentioned earlier, the majority of online shopping activity starts on a different site from ours. With such low barriers to entry and the ease with which consumers are able to move from one to another, we strive and work incredibly hard to ensure that we are the search engine that people come back to. Hugh’s point earlier that the dynamics of the online environment can shift in a matter of months is worth dwelling on because that is how consumers will use us. If we are useful they will hopefully keep returning.

The Chairman: Rebecca, do you see yourselves as a dominant utility?

Rebecca Stimson: No, I do not think so, and I would reflect similar comments, in that if you think about things people use Facebook for, such as messaging each other, sharing photographs, looking at news online and so on, there is an enormous range of other companies and platforms that will enable you to do those things, so, no, I do think I would recognise that.

The Chairman: Nonetheless consumers do. Baroness Chisholm.

Q178 Baroness Chisholm of Owlpen: Do you think there should be a public interest test in mergers between businesses which rely on user data?

Katie O'Donovan: I am not an expert in what already exists in terms of public interest tests in mergers. I believe we have a very robust system here, but I would need to get back to you on that one.

Rebecca Stimson: As I said in my previous answer, the competition regulations here are very stringent and thorough, and I am sure that kind of test must exist. I am not an expert in it, but these kinds of mergers and acquisitions are happening in this marketplace under the full scrutiny of the current UK regime.

Lord Gordon of Strathblane: The point is if you take media mergers, there is not simply an economic test; there is a public interest test as well as to whether it is a good idea for society that it should happen. The simple question is: would that be a good idea in the field of your companies?

Hugh Milward: In which market? Part of the challenge is to define what the market is.

Lord Gordon of Strathblane: I accept the problems but, on the other hand, we have been told before that Google has a 94% market share in the UK.

Katie O'Donovan: I do not think that is accurate. I think Microsoft would challenge that statistic.

Hugh Milward: We have at least 12%.

Baroness McIntosh of Hudnall: I want to reframe this question, if I may, Chairman, about the utility issue, because each of you answered as if you were being asked whether your company is a utility. That is not really the question. The question is whether access to the internet is a utility. It is quite obviously the case that electricity is regarded as a utility. There are many companies operating competitively within that market, but it is regulated as a utility for reasons to do with public interest and the public good. Could you very briefly answer the question again in relation to access to the internet as analogous to access to clean water or electricity or any other utility you can think of?

Katie O'Donovan: In the UK, access to the internet is often provided not by the companies here today but by telecommunications companies, and I believe there are statutory duties on them. I am not looking to obfuscate and avoid answering your question, I am just saying that we are not responsible for the access to the internet in the UK.

Baroness McIntosh of Hudnall: I am not asking what you are responsible for; I am asking about an issue of principle and what your view is.

Katie O'Donovan: The Government themselves have said that access to the internet is of inherent value to UK citizens. They have their own standards and expectations for the speed at which that should be delivered and the availability of that in rural areas and elsewhere. As a company which relies on internet access, that is absolutely welcomed. If you are asking more broadly if should there be regulation in this space, or of what our companies do, that is a slightly different question, which I am very happy to get into a discussion on.

The Chairman: If only we had the time. Do the other witnesses wish to add? No. Baroness Bertin.

Q179 Baroness Bertin: I should first declare that I work for BT. I would like to talk a little about content and user-generated harms. I would kick off by asking what responsibilities do you have in terms of moderating user-generated content?

Rebecca Stimson: Obviously some of this is covered by the existing regulatory framework, including by the e-commerce directive, where we have a liability, as I am sure you are aware, for illegal content online. There are other things that apply to content such as data protection, GDPR and so on. As I am sure colleagues here have, we have a broader sense of where our responsibilities lie. That is best demonstrated by our content standards, which are very extensive documents on what we do and do not allow on the platform, which are public, so that people can see and understand those. They concern a whole range of different issues, from things such as sexual exploitation images, to bullying, terrorist content, nudity.

Baroness Bertin: If I can stop you there. You know all the figures—the NSPCC has published them—and we still have terrible figures. Some 25% of children have seen content on Facebook and YouTube that contains suicide, and I could go on. Something is not quite right yet, is it, especially in terms of protecting children online?

Rebecca Stimson: I would recognise that there is always more we can do. I am not going to sit here and say it is all fine because it clearly is not. The statistics show that the overwhelming majority of people who engage with Facebook have a positive experience and see good things but, while there are people who see bad things, there is more for us to do.

Baroness Bertin: Sure, but it is about priorities, is it not? It is important to acknowledge that as an industry you have come together and done good work on terrorism, for example, but—and correct me if you think this is wrong—from speaking to law enforcement agencies, it feels that the child protection element is a rather hard yard, if you do not mind me saying. I would love to know how high up your list of priorities this issue is in your companies, how much time is spent on it and how much brain bandwidth you are giving to putting in new ethical designs to change the agenda, so that, for example, Katie, if I were to put in a multi-layered search I would not get category A child abuse images, which you still can get, I understand?

Katie O'Donovan: Can I start on answering that question? We have a very clear policy against any illegal child sex abuse material being discovered through search. I too have talked frequently with law enforcement agencies and have said to them that our algorithms are set not to return any of that content.

Baroness Bertin: But, as I understand it, you could do a repeated layered search and still get category A child abuse images.

Katie O'Donovan: I do not know what “repeated layered” means.

Baroness Bertin: This is the language the law enforcement agencies use and they say they have evidence you could still get those kinds of images.

Katie O'Donovan: I have talked very frequently to law enforcement agencies and made clear that our policy is not to deliver any content through search that is classified as child sex abuse, and, if they have information or instances where that is not working, to please let us know so that we can ensure that our services are working as well as possible.

To answer your broader question about how we collaborate on issues such as this, we are members of the Internet Watch Foundation, which is an organisation based in the UK. As I said earlier, I am on the board. It is a world-leading organisation, with over 100 members, technology companies, big and small, working together to tackle the issue of illegal child sex abuse imagery. For Google itself we have developed technology that not only uses hashes, which have been developed by Microsoft and others to identify known child sex abuse, but just six weeks ago, we announced that we have developed a review classifier which increases significantly the effectiveness of our human reviewers looking for this content for the first time to identify unknown child sex abuse imagery. It really is treated very seriously.

If we move on from illegal child sex abuse imagery, I think you also asked about issues of concern to young people, and, obviously, young people online can come across content that they are not ready to see or they do not want to see or it is inappropriate that they see, and, as well as having community guidelines on YouTube video-sharing platform that go above and beyond the law and are enforced through our flagging systems using technology, we have also invested really heavily in YouTube Kids, which is a platform for under-13s to access some of the user-generated content and some of the content people really like from YouTube but in a much safer and more relevant environment.

Baroness Bertin: To build on that, in terms of your investment and putting your not inconsiderable brain power into these issues, what percentage of R&D do you invest? Do you think it is high enough up the agenda in your boardrooms?

Katie O'Donovan: From the R&D we put into this, we know this is one of those areas where artificial intelligence can be hugely beneficial. We are able to use artificial intelligence that has been developed for a general purpose to help us identify content that we think may be child sex abuse imagery. In that case, that is changing the way we have been able to tackle this issue.

Baroness Bertin: Do you have a figure for your R&D?

Katie O'Donovan: The R&D that we use is developed for general purpose so we do not have one technology team here that works solely on this issue and another technology team over there. There is technology that is developed for a general purpose which can be utilised in different ways. That is a benefit to us. That means we can use software that is developed through commercial means to help us on these issues. We make sure we do because we realise we have the resources to invest a significant amount of money.

Baroness Bertin: You could completely change the game and solve a lot of these issues, I would have thought.

Katie O'Donovan: I would not necessarily claim that we could do that, but we have been able to invest really significant amounts of money and computational power. We developed video-hashing technology, which enables us to find content of videos which is known child sex abuse imagery. The development of that technology has required significant resources and investment and we are making it available to companies big and small. Microsoft and Google collaborate often on this topic, but we go beyond that to companies which could never begin to afford this technology. We take this really seriously. It is an incredibly difficult issue which we know requires an industry-wide response and the right policies and investment in technology, and we absolutely have made that.

Baroness Bertin: Obviously, you would never dream of selling a product that did not have anti-malware and antivirus programs. How much ethical design are you putting into your products? How much more thought will you be putting into horizon scanning going down the track?

Hugh Milward: There are different aspects of ethical design. For example, building accessibility into the fundamentals is fundamentally important. Going back to the drawing board of the use case of a particular product, it is about how you make sure that it is accessible to everybody. That is an ethical way of building a product. Hopefully, that is a given. It is not just about ethical design; it is also about how much effort is going into the development of these kinds of things. One of your questions a little earlier was implying that surely we can solve this if we put enough effort into it. The risk is that we lull ourselves into believing that is possible, but what happens is it just gets driven to the dark web and it becomes out of the control of all of the companies sitting at this table, and we have no control over the dark web. That is what happens.

All the companies here use PhotoDNA, which we spent a lot of time developing 20 years ago. We have just concluded a massive engineering project on VideoDNA. It was launched a couple of months ago, and we are hoping to use it in exactly the same way. It requires significant amounts of engineering effort to do that. It will produce cleaner and cleaner results the more we use it. The behaviours of those predatory paedophiles are not being addressed through the actions of those at this table. Their activities are being driven further and further to the fringes of what we have control over, and that means the dark web.

Baroness McIntosh of Hudnall: Shall we move on to the moderation point? It comes back from the issues of the dark web, which are outwith your control and, indeed, anyone’s control at the moment. When you look at content uploaded on to your sites, you have ways of moderating it that include algorithmic methods, and, I imagine, humans. It has been put to us by more than one witness that the number of human moderators who are actively engaged in looking at content on your sites is very small compared with both the number of users and with the amount of algorithmic moderation that goes on. Do you think the balance is right between those different kinds of moderation and do you have any plans to extend them in any direction?

Rebecca Stimson: You are right that at the moment it is a balance of automated moderation and moderation by humans. The automated systems, as has been alluded to in the previous answer, are really good in some respects, such as for detecting terrorist material and child sexual exploitation. Some of the statistics that all of us are producing are very good. Recently, similar to Google, we announced our new tools for unknown child nudity images. They have an extremely high success rate. There are some more complicated areas such as bullying which still require human moderation. Sometimes it is about context and sometimes it is difficult. While we have a certain amount of automation to spot it, it often still requires a human being to review it. As you say, we are investing very heavily in the machine learning but it has to be accurate. We cannot have it misunderstanding what is happening and censoring large amounts of content unnecessarily. We have gone from 10,000 to 20,000 people working on safety and security in Facebook.

Baroness McIntosh of Hudnall: Is that world wide?

Rebecca Stimson: Yes, that is a global figure.

Baroness McIntosh of Hudnall: How many users do you have?

Rebecca Stimson: We have around 2 billion users. That is alongside the tens of millions of pieces of content that machine learning is able to look at in several of the most important areas.

The Chairman: Can you give the number of users in the UK? You have given the global number of users.

Rebecca Stimson: I believe the number of users in the UK is 40 million.

Baroness McIntosh of Hudnall: And how many moderators would be looking at that area?

Rebecca Stimson: The way that moderation works is we have teams around the world 24 hours a day, seven days a week. When people report content that requires moderation, there is not a UK team to look at UK content; it will be sent and will depend on what it is and whether it needs specialists to look at it or whether it is a matter for law enforcement. I am not able to give you the number of UK moderators for so many users.

The Chairman: Something as context driven as bullying, for example, probably needs a UK moderator to properly comprehend it. Is that right?

Rebecca Stimson: It depends. Hate speech is an example where more local knowledge can be helpful. We call them flows and that is where the different pieces of content being reported go to different places. I am afraid I could not give you a specific number of users and moderators in the UK because it does not work that way.

Baroness McIntosh of Hudnall: It would be quite helpful if you could give us some idea, given that, exactly as you say, context is everything with some of this stuff, and language is also very important, not just whether it is English or French, but whether it is English used in a different way. English is used in a different way here compared to America or Australia. If you could give us some notion of how local to the UK moderation and investigation of content on your site is, that would be very helpful.

Rebecca Stimson: I am very happy to try. As I said, it is not quite how that system works but let me come back to you with an answer.

Viscount Colville of Culross: We had evidence a few weeks ago about the number of moderators that you have in Germany compared to the number of moderators you have world wide. There is a hugely disproportionate number in Germany because of the law it has about hate crime. Is that not tempting legislators here to try to follow the same course to make sure you have more moderators?

Rebecca Stimson: I think you are referring to the netzDG law.

Viscount Colville of Culross: I am indeed.

Rebecca Stimson: I will take a moment to be clear about what that law involves. There are hate speech laws in Germany which almost entirely map across to our own hate speech rules whereby if something is reported to us we take it down. Some things in Germany are specific to Germany. The netzDG law required us to introduce a reporting mechanism so that people in Germany could report content under that law. In response to that, we designed into our platforms a way of reporting under that law. We have moderation centres in Germany that were there before. There are more people working in those centres now but, as I said, those centres operate in a global way, and that is part of the general increase in security and safety personnel working within Facebook.

It was not as a direct response to that law. As a direct response to the law, we created a way of reporting, and we took on more lawyers, because what is interesting about netzDG is that it sets quite a tight timeframe for companies to decide whether a piece of content is illegal or not. Sometimes that is very obvious but sometimes it is not. Sometimes it is more of a fringe case of the sort we were just discussing. I know there is a very live debate in Germany about some of the consequences of that law: for example, the risk that it might incentivise people to err on the side of caution and take things down more liberally than they might have done before. We introduced a new reporting system and lawyers. Our content moderation centres in Germany deal with global content and are not a response to that law specifically.

Viscount Colville of Culross: Is your view that the law is encouraging people to take material down which would not be taken down by the legislation?

Rebecca Stimson: All I am aware of is the debate that is happening in Germany. I am not in a position to describe whether that is actually happening. I know that there are a lot of concerns about an unintentional perverse incentive that that law may have encouraged. As you know, very significant fines can be imposed on companies for not removing illegal content quickly, so you can see that where there is what we call an edge case, you would tend to take it down, rather than, as the clock is ticking, spend your time debating whether it is, strictly speaking, illegal. I am aware that it has been quite a controversial piece of law in Germany.

The Chairman: Do you have anything to add?

Katie O'Donovan: I would add that the reviewers make timely decisions about whether the content stays on the platform or is removed. One problem we have heard about from lots of different people over the last couple of years is the scale of the decisions we make, what those decisions are and the timeliness of them. In spring of this year, we started to publish a quarterly transparency report on the content on YouTube that is flagged. We detail it by the category area that it is flagged under and provide information on what happens with the content that has been flagged. That is an iterative process, so each quarter we have been adding more information to that, and we will continue that, based on the areas that people are interested in. To go alongside that, we have also published a user report history for users. If you flag content on YouTube and you want to know what happened to it, you can go on to your report history page, and it will tell you if the content is still live or it has been taken down. Again, that helps people evaluate whether the system is working well at a global level, through our transparency report, or, at a personal level, through the report history.

The Chairman: Let us move on to platform liability. Lord Bishop.

Q180 The Lord Bishop of Chelmsford: This follows on from the discussion we have just been having, looking not so much at how you moderate the content when it is there but who is liable and responsible for it being there in the first place. The first question is simple: to what extent should online platforms be liable for the content they host? I am particularly interested because in our previous inquiry, Simon Milner from Facebook gave evidence to us and conceded that Facebook was something in between a publisher and a mere conduit and therefore perhaps there should be some additional liability, particularly in respect of advertising. I understand that you continue to use the word “platform”, but what is your understanding of that in terms of liability and, perhaps more generally, responsibility?

Rebecca Stimson: Simon recruited me so obviously everything he says is perfectly accurate.

The Lord Bishop of Chelmsford: But perhaps in this case uncomfortable.

Rebecca Stimson: We have already touched on the statutory responsibilities we have currently under the e-commerce directive and a number of other codes and standards that are applied to us. That tends to apply to illegal content, and, as I said in the previous answer, we broaden that out to a whole range of ways in which we consider ourselves responsible for the content of that platform to ensure that what people are seeing is not harmful, it is not hate speech, bullying and so forth, or containing fake adverts, for example. That is clearly illegal and, again, we have a responsibility under existing law to remove that kind of content. You can take it into fake news. We have done quite a lot on the platform over the last year or so and have removed nearly 500 million fake accounts. We have changed the algorithm that underpins how Facebook works to remove as much fake content as we can. The issues are very big, and there are a huge number of tools in the tool box we can use to deal with them, which reflects the fact that we take a very broad approach to our responsibilities for what is on the platform.

Some people talk about extending the principle of content liability into other areas, and it is worth reflecting on what that would mean if you started to go outside what is illegal and make companies liable for pieces of content that are not illegal but are harmful. I know from our very useful conversations with DCMS, which is thinking about this in the context of the Government’s forthcoming White Paper, that, clearly, if you get into that kind of territory, you need to be very clear what the harms are and be very specific about what you are talking about. If you are going to declare something illegal that has not been through both Houses of Parliament but is in some other category, you have a slight risk of confusion and inconsistency there, and you need to be very clear.

The Lord Bishop of Chelmsford: But you can see the problem.

Rebecca Stimson: If you think about the harms you are trying to address and look at the particular successes that we have mentioned in the last couple of answers around how effective self-regulation is in some of those major harms, and the statistics we can all give about how much of this content we are successfully removing, you need to think about extending that principle of liability to a company as well as to the host of the content. It is user-generated content. I think you can see there are some complications there. I am not saying it may not turn out to be a good idea but you have to work from a first-principles basis.

The Lord Bishop of Chelmsford: What do you think should be done to address that issue where you can recognise the harm in the content? It does not seem to me sufficient to say, “We are just the platform”.

Rebecca Stimson: The way we operate at the moment, and the way that the Government have been consulting on this, is to look at the policies that we all have at the moment, and their transparency and the reporting on them, to see how well they are being implemented, and we are held to account and scrutiny on that basis. It gets very difficult because all our platforms work in different ways. They have different technologies that underpin them and they serve slightly different purposes. A narrowly drawn prescriptive liabilities law could be quite difficult to implement in practice, and, as we have talked about with netzDG, could have unintended consequences. There should be a broad framework of principles for what we are expected to do, with an emphasis on the transparency, for us to show you what we are doing, and for you to hold us to account.

The Lord Bishop of Chelmsford: That would be helpful.

Rebecca Stimson: We all produce extensive transparency reports and, as you know, that is a big part of the Government’s consultation that we are all collaborating on.

The Lord Bishop of Chelmsford: There are some specific issues to do with advertising, but I might let other colleagues come in on that.

Lord Gordon of Strathblane: May I come in on this because it is directly related to the point the Bishop was making? I was going to ask Katie about this. In your evidence, you say in praise of the e-commerce directive that it ensures that those who post material online take responsibility for the content that they produce. The great problem is they do not, frequently, and they leave you with the problem of taking down the offensive material subsequently. Would it not be in your enlightened self-interest if it was not put up in the first place? What steps could we take to help offensive content not appear in the first place, without saying that it is your responsibility?

Katie O'Donovan: I think that is a good synopsis of the conundrum facing us. We want to have an open platform. In the vast majority of cases, it is used in a wholly responsible way. We absolutely see our part in ensuring a responsible framework for the hosting of user-generated content. There has to be some personal responsibility and certainly in YouTube that can be impactful. We have very clear community guidelines and if people breach them, they know their channel can be removed if they do that persistently. That is a serious penalty before you even reach the law. There has been a vast increase in the number of individuals who have been prosecuted for online hate speech or associated crimes in the UK. I think people are beginning to realise that the internet, as Hugh said earlier, is not the Wild West. An element of personal responsibility is key, but it does not need to be this conversation between the e-commerce directive or publisher; there is a balanced situation and a balanced ecosystem that is emerging within that.

Lord Gordon of Strathblane: To go slightly further, do you think it would make your job easier if people were not allowed to post things online unless they had an identifiable traceable address and there was some procedure for seeking redress if something was wrong?

Katie O'Donovan: The question to ask is not necessarily what would make our lives easier. It might make particularly tough questions easier to answer, but we need to think about the detriment to all the law-abiding responsible users who want to upload a video they made in their shed where they have created a model steam engine. The vast majority of content we see on our platform is completely innocuous.

Lord Gordon of Strathblane: And nobody wants to do anything about that. We are talking about the content that you are subsequently asked to take down because it is offensive. Would it not be better if it did not reach you in the first place?

Katie O'Donovan: But how do we stop only that content reaching us and not the positive content?

Lord Gordon of Strathblane: I accept that there are different points of view on anonymity, and anonymity is a benefit in some regimes, but, equally, it means the system could be abused by people who are untraceable.

The Lord Bishop of Chelmsford: To pitch in, you could do some moderation before the content goes up rather than afterwards.

Katie O'Donovan: I have been involved in conversations around anonymity for many years, and I can understand why it sounds like a good solution. Simon gave evidence to a different committee where he said that Facebook requires a real-name policy, and we have certainly seen some issues there. Before I worked at Google, I worked at an organisation called Mumsnet, which is an online forum for women. There is a policy of anonymity on there. People can choose their own names and because they have that anonymity they are able to exchange stories about domestic violence. When I was at Mumsnet, and subsequently, it came under online attack from men’s rights activist organisations which sought to, and did in some cases, illegally share the data of users. It is very difficult to have a system which ends anonymity and does not end the right for people to be able to have really difficult conversations online. The peer-to-peer support that has been enabled for whether it is domestic violence victims or on other issues where people are looking for that element of peer support, is really important. I do not think there is an easy way to end anonymity for the bad guys but keep it for the good guys.

Lord Gordon of Strathblane: Do you think that abuse of the system by the bad guys is a price worth paying to preserve it for the good guys?

Katie O'Donovan: I do not think tolerating abuse is worth it, and we do not tolerate abuse. An open internet for us does not mean a free for all. We abide by the law and everyone who uses our platform has to abide by the law. We also have our own community guidelines that go further and we enforce those and people will be removed from our platform if they break them. It is difficult, complicated and resource intensive but for us it preserves the free internet.

Lord Gordon of Strathblane: Could I ask a further question of Rebecca? A point was made about whether you are a platform or a publisher, and one agrees that it can be a somewhat sterile discussion. Are there objective criteria that would determine where on the spectrum you are, or is it a self-defining matter?

Rebecca Stimson: I can totally understand why this debate is happening. As I said, we have a range of responsibilities, some of them statutory and some of them we assume ourselves for the content that we carry. I can completely understand why Simon—and I do not know if he was in front of this Committee or perhaps a different one—said there is probably a third space in which our responsibilities lie. We do not have editorial boards or teams of journalists, as you have just been talking about in your previous question. We do not moderate content before it goes up. We are clearly not a publisher, but our responsibilities towards the content are a matter of very lively debate around the world.

The Lord Bishop of Chelmsford: I will come in very briefly because it seems to me—and I know I am sounding like a scratched record—that we should be having precisely that debate, and asking whether there is a different category, a different way of defining and describing what you do. From that it might make it that little bit easier, both for you and everybody else, because, frankly, a lot of people get fed up when we hear, “It’s nothing to do with us. We are just this wonderful space which people occupy and we can’t control it”. That is deeply frustrating, which I know you know, but if we put the work in, do you think it might bear fruit? I know we cannot do it now.

Rebecca Stimson: We definitely would not say it is nothing to do with us. I have been quite clear, hopefully, about the responsibilities we feel towards content. I think you could have this debate and it may be fruitful. The approach we are seeing the Government take in the White Paper is that we need to think about what harms we are talking about. That is the focus. Whether you want to label us as something else, a third thing, perhaps that will emerge as the answer to addressing those harms. The approach they are taking is to ask what are we really worried about, where is the underpinning evidence and analysis and what is the best way to address those issues, be that different kinds of regulation, codes of conduct, all the various things we have been discussing this afternoon. It is a slightly second-order issue to work on our definition of what we are, unless it becomes apparent through that process that that is the key to addressing the harms.

The Chairman: May I ask you—because you have implied you might be some sort of third thing—have you defined what that is? Have you discussed internally what that might mean?

Rebecca Stimson: We come at it, as I have just said, from the approach of the harms that we are trying to solve.

The Chairman: No, you indicated and repeated that you are possibly some kind of third thing. Have you had a discussion about what you mean by that?

Rebecca Stimson: We have not had a discussion about defining exactly what that might be because, as I say, we are looking at the debate through a different lens, which is what harms are we trying to tackle and what is the best way to do that.

The Chairman: Should we move on? Baroness Kidron.

Q181 Baroness Kidron: I was really struck by something in all of your opening statements, and forgive me if I paraphrase you, but Rebecca said it is really hard to keep abreast of the pace of change, Katie said the ways in which people use technology in their daily lives is changing and Hugh was talking about the question of trust. What struck me about those statements was it was as if you were not engaged or not the motors or not responsible for those things, for that journey.

My question is around design of service, not necessarily about this list of harms we have gone through, but other sorts of ways in which you are pushing the world order. I am randomly grasping this from the air, but one example might be that the vast majority of YouTube videos are watched as recommended by YouTube algorithms. The vast majority are in a loop where you watch one, and it is offered up and you watch the next and the next. Another example is Facebook’s decision to have friends of friends on its Messenger service, so, even if you are under 13, your world can extend to friends of friends. I do not want you to hook on to the particular examples. You know from the IWF that we have seen a huge increase in child abuse images, and we talked about that just now, and we have talked about mental health, but we have not talked about what we are seeing in terms of compulsive use and how design of service encourages compulsive use. We have not talked about the fact that spreading data is very difficult. I would like you to talk a little about what you feel your responsibilities are in the design of service that is not about content and those kinds of harms, but pushing the direction of travel, and how you feel about the fact that when you get on the bus, if you get on a bus, every single person on the bus is going to have their phone at their nose. Let us start in a different order with Hugh. He is a bit safer on this particular issue, but please carry on.

Hugh Milward: We are consciously developing technology that will make people redundant. What is our responsibility in that? We know that the pace of change in artificial intelligence, first, is causing people to have fears, and they are right to have fears, and, secondly, they will lose their livelihoods as a result of it. Does that mean we should stop developing it, or does it mean that we should step forward very carefully and design interventions that help to mitigate some of those fears as we go? We have never done this before. The technologists who designed the plough did not think about the impact on those affected by the consequences of that development. This is a new thing that we are bringing closer and closer. We are shrinking the gap between the design of the technology and the design of the mitigating interventions. This is extremely welcome and where we need to go on this. We will not get it right. There will be use cases for technology that we cannot predict now that people will be concerned about. We are taking more, bolder, clearer, more consultative, more collaborative steps in the way we design them now than we have ever taken before.

Rebecca Stimson: I would reflect on a few things. As you know, about a year ago, we made a major change to the algorithms that underpin Facebook to move into more meaningful interactions, to ensure that people were having a better experience online and that there was not so much fake news and clickbait and so on. A large body of evidence suggests that has been very successful. Recently there have been three studies in the US that show that the levels of people engaging with fake news, for example, have dropped by 50% in a year, which is really great stuff. We have also touched on some of the incredible advances in AI and machine learning in addressing some of the worst harms.

Slightly reflecting a previous question, some of those technologies are now so effective at spotting that content, it is almost instantaneous that it is able to take that down, and nobody sees some of it. It is not quite the same as pre-moderation but it is split-second stuff.

We are also all members of lots of global consortiums that are developing technologies, as Katie said in a previous answer, that are useable by smaller platforms. When we get into a conversation about harms, the large platforms are pretty transparent and open and we are held to account. Certainly in our work with the Home Office, they tend to be much more worried about—

Baroness Kidron: To be clear, I am not particularly talking about harms. I am talking about the societal piece.

Rebecca Stimson: The reason I was mentioning them was that we are sharing some of the technology that has resulted from our R&D investment with smaller platforms, to ensure that we are not just hoarding that kind of technology to ourselves. We have a centre—I believe it was established this year—for AI ethics within Facebook, which is looking at these very complicated challenging issues, and we are fully participating in those debates.

Katie O'Donovan: It is a really good question. I agree with the way Hugh described it. The gap between the technology being developed and the mechanisms to help us maximise the potential of that technology, to put it positively, is much smaller. Google, like all big tech companies, has an annual developer conference. It is called Google I/O and is akin to a party conference for tech developers, where you get together and show the brightest and best of the work that you are working on. You have limited time to get stuff in because it is such a high-profile event. Sundar Pichai, our chief executive, spent a significant amount of his presentation talking about the technology that we are developing to help with digital well-being.

You and I have talked about this before. We all find our mobile phones particularly helpful, but we also find that we spend too long on them. The technology we have developed helps people understand what they have spent their time on their phone doing each day. You can set a timer for a particular app, so you might allow yourself to be on email for longer but on social media for less time. It tells you how long you have spent on video platforms. Again, we have adopted some of that technology into YouTube. You can turn off auto play on YouTube. You can find out how long you have been watching YouTube and set a timer. Again, we are developing technology specifically for families and younger people. It is a great question to ask because it is exactly where we should be investing our resources to ensure that technology is a tool that we as people choose how to use and on what terms, and make the most of it that way.

Baroness Kidron: To that point, you mentioned resilience and the amount of money that goes into schools, but one of the things I struggle with is this idea that we create technology that is very problematic for people and try to make them resilient to it rather than we create technology that is really good for people. Even the Time Well Spent movement and the wellness thing is picking up the pieces at the end. I would really appreciate your answers here. There were 17 industrial Factory Acts. There is such a bottom-line issue here and I know you have said you are just little people who are in competition with the others, but look at the share price and your position in the market and think how can we, with your interests, put you in charge of what ethics looks like? How can we not take a more societal view about the development and design of services?

I want to hear from you, but, to put on the record, we have Tristan Harris saying that the technology has hijacked our psychological vulnerabilities. We have John Naughton saying that the future looks pretty bleak because we have a business model of surveillance capitalism. Doteveryone says the design processes and business models of technology need to be strengthened and regulated. I am not going to bore the room, but I have another 12 on this list. I want to ask the question in a slightly more robust way.

Baroness McIntosh of Hudnall: May I add a sentence to amplify what you have just said, in a rather less creative and, you might think, more hostile way? All those interventions you describe, as Baroness Kidron says, are ways of mitigating the problem rather than preventing it, so if somebody switches off, they do, but if they do not switch off, they are still in the world of that particular kind of behaviour. What if all of your interventions were really successful and all the people to whom you were offering the opportunity to mitigate the potential damage that might be done if they go on using those apps, what if that worked, what would that do to your bottom line?

The Chairman: If you could answer this in the round and we will move on.

Baroness McIntosh of Hudnall: I wanted it to be part of the same question.

Rebecca Stimson: I do not think we consider that ethical design has been outsourced to us. I have already mentioned several things that we do within Facebook to ensure ethical design. It depends what you are talking about. It can be anything, from whether our terms of service are clear, accessible and understandable to people, to how we deal with younger users, to algorithms. What we are talking about specifically will depend on the right kind of ethical response. We have a number of ways in the company of doing that, but, as you saw in the Budget yesterday, the Chancellor announced more detail on what the Government’s own data ethics centre, I think it is called, is going to be doing. They have announced some really interesting initial work that we look forward to working with them on. We do not feel solely responsible. All the organisations at this table partner with hundreds of organisations around the world, focused on everything from algorithm ethics to child safety and so on. Ideally, we try to address harms before they happen, but where they have happened, we try to stop them happening again. I know you are aware of the many programmes that Facebook runs.

To answer your question, when we changed the algorithm that prioritises what you see on Facebook, our chief executive was very clear that we would take a hit on the bottom line, and we did. We have seen around 50 million fewer hours spent on our platform. We wanted to do that because we wanted it to be a long-term positive and useful product in people’s lives; a product that they enjoy and is good for them to contact their friends and family through. It is not in our interests to have it be a terrible, addictive and unpleasant place to spend time, so we have taken a hit on our bottom line and seen less engagement as a result of changes we have made consciously for that reason.

Katie O'Donovan: To build on that, one point I would like to make is that not all screen time can be treated in the same way. People use technology in very different ways. It is good that the Chief Medical Officer is looking into that from a UK point of view to find out what more needs to be done on that. In terms of our bottom line, the way that we operate as a company is that we deliver to people the information that they are looking for. If you look for something on search, you do not want to spend a long time on Google; you want to go through to it. We have built the products because we want them to be used and we want people to be able to manage how long they spend online in a way that works for them.

Baroness Kidron: Katie, I was not talking about screen time, and I absolutely agree with you that not all screen time is equal, but a lot of the design elements have factors and push factors that are not necessarily in the best interests of the person, or at least are somewhat determined by your algorithms that may have stickiness or other things that they want to do.

Katie O'Donovan: That is where we need to clearly define the issue in question and what is needed to be done from a technological point of view.

The Chairman: Who should define? You said “we” need to clearly define.

Katie O'Donovan: The new technology that we announced in the spring to help with digital well-being covered everything from the amount of time you spend on your mobile phone per se, to the amount of time you spend in different apps. Some are enterprise or work apps, some are educational apps, some are multi-purpose apps, where you could be on a social media platform doing something very flippant or you could be on a social media platform contacting your friends and family. For us to understand, we believe that the technology that we announced at our I/O helps users and puts them in control of how much they time they spend online, limits the amount of time they spend online and gives them information about that. They can turn off notifications and bundle notifications. There is a new feature whereby if you put your phone screen down, you will not get any notifications at all. We think all those things are positive innovations in this space, but if they do not go far enough, or if there are further requirements that we would not choose to invest in in technological terms, and society requires of us, it is appropriate for the Government or for your committee to make recommendations in that area.

Baroness Kidron: That is why I am coming back to this idea of outsourcing, because a lot of these things have been responses to various forms of pressure, either in advance of threatened regulation or as a result of regulation. That is really why I am raising the issue. You have all said in answer to the first question, “We are doing rather well and we are doing our best and making these big investments”, but is it reasonable to leave you to choose where to make those investments and decide what those boundaries are, or is it not up to society more broadly? I am sorry, Hugh, I interrupted you.

Hugh Milward: The Warnock commission is a very sound model of this. The situation was you had an advancing technology, you had a segment of society that saw the technology almost as its salvation and you had other segments of society that were deeply concerned about the development of this technology. You bring in the country’s foremost philosopher, who gathers a group of big minds, and those who are developing the technology, and create an ethical framework by which the development of that technology is guided. That is a very sound set of principles by which we can take forward the development of technology in the UK. I am extremely encouraged by the approach the Government are taking at the moment over the AI Council and certain other aspects. We are finding ways of stepping forward together in the right way, in a way that is not outsourcing those ethical decisions simply to the technology companies.

As we build ethical designs and behaviours into the way we work, we will suffer at the bottom line. That is fine. That suffering at the bottom line will not least be because other cultures that take a different ethical stance or perspective on the way that technology should be used will advance. They will sell into markets and to customers that we will not, and that will mean that we will not be as financially successful. There will be a penalty for that, and we are fine with that.

The Chairman: I think we have bottomed out to quite an interesting issue at the heart of this as to whose responsibility it is to guide you societally, on top of the work you are doing as you develop products, and that is an interesting focus. I thought Mr Milward’s analogy was interesting. Sadly, we need to move on. Lord Gordon.

Lord Gordon of Strathblane: To segue from that subject on to another, you are almost arguing for a superregulatory supervisor, which will ensure that self-regulation or co-regulation is working, or point out where it might need to be statutory. Were you arguing for that? I thought you were.

Hugh Milward: It is less about a supervisor and more that, as we take ethical decisions, those decisions cannot be divorced from society. We need to find ways in which we are making sure that they are aligned and consistent and that we are not just bypassing the will of this House and the other.

Lord Gordon of Strathblane: In long-term self-interest, it would be good for everyone if public interest and self-interest were aligned.

Hugh Milward: That would be a marvellous nirvana.

The Chairman: Shall we move on to the final question?

Q182 Lord Gordon of Strathblane: Turning to GDPR: what do you think are its strengths and weaknesses? In answering that, all three of you might pay some attention to what Tim Cook of Apple, which I understand has a very different business model from yourselves, said about privacy and everything else on 24 October in Brussels?

Rebecca Stimson: The GDPR established some very important principles, some of them reflected in what we have said today, at the forefront of that being that a user’s data is their own and they need to be in control of what happens to it. They need to give clear consent around how it is used and there need to be very clear rules for the use of their data. A harmonising-piece legislation across all 28 member states to drive up standards is an excellent thing.

There are some complications in it, as I said in my first comments. One of your previous witnesses said it was a once-in-a-generation reset of data protection law. I know from my experience as a civil servant looking at it that it is difficult to write anything in this space which is totally future-proof and reflects everything. I remember that there was an interesting debate around rights and expectations of rights. There was a particularly controversial debate around the right to be forgotten and the right to delete. There is a challenge around setting up a right in people’s minds and, in practical terms, whether it is possible to delete things once they are out on the internet. It definitely has strengths and weaknesses. It is very early days—it only came into force this year—to judge fully.

Katie O'Donovan: When you ask that question, it is important to ask that question of us, and I will answer it from Google’s perspective, but it is a piece of legislation that has impacted almost every organisation in the UK that handles any sort of data. It is worth full and holistic scrutiny. For us, the aims and ambitions of the regulation are exactly as we see them: to give users control over their data and transparency over how it is used. There were a couple of things that we were able to do in advance of the GDPR which very clearly related to the direction the GDPR was going in.

We did some of them because we were able to do them well ahead and before they were instituted in the GDPR, and I can go into a little bit of detail. The information that we provide through My Account, which I mentioned earlier, is really meaningful information, whereby consumers can understand how their information is used by Google and decide what they want to share with us and what they do not. That is really positive and in the spirit of GDPR.

The other thing we have long had as a company—and GDPR now requires other companies to do—is the ability to take out your data. If you have a Google account and you use Gmail, our email provider, you might accrue an enormous number of emails, some photos, your contacts, whatever else, but we wanted to make sure that you could take your data to any other provider, so we have had a system called Takeout, where you can remove your data from Google and take it to another provider. That has been instituted through GDPR and we think it is really positive.

As Rebecca mentioned, having a single set of standards across a large number of people is really helpful. As the UK moves toward Brexit, we hope there is continued data adequacy between the UK and the rest of Europe.

Hugh Milward: In the US, we called for privacy legislation back in 2005, so it is no surprise that we decided very quickly to adopt GDPR as the benchmark for privacy globally for our company. We pretty much treat it as the gold standard in data privacy. What is interesting is that the tools to help manage the privacy settings for our customers that we built on the back of GDPR are now in use world wide. Some 400,000 of those privacy settings, so the second-highest number of users, are in the UK. Interestingly, the highest number is in the US where GDPR does not apply. It is early days and the UK is one of the few European markets that has adopted GDPR in its fullest sense and as early as it has. There are a lot of European countries that have not yet got to the stage that the UK has, irrespective of Brexit. It is still early days but we are very positive.

Lord Gordon of Strathblane: I commend all three of you for being too polite to make any criticism of Tim Cook of Apple, a fellow member of FAANG, because he will be giving evidence next week, and presumably would retaliate. Could we look at how we might make the average user more aware of what you do with the data? Would it be an idea if a little icon lit up on your screen when your data was being collected? Would you object to that?

Katie O'Donovan: Again, it depends what you consider to be data. If you were doing a search on a search engine for Wellington boots, by typing “Wellington boots” your data is being collected. We want to help users to understand what data is and how we use it. We have advertising for My Account across Google, across the search engine at various places where they can do that. We certainly work very hard to communicate how it is used.

Lord Gordon of Strathblane: But you would have no objection to the principle of an icon lighting up when data was being collected?

Katie O'Donovan: I am not a user interface designer but we want users to have more understanding of where their data is being used and how, and have more control over that. That is absolutely how we are building our systems.

Lord Gordon of Strathblane: I see Hugh nodding.

Hugh Milward: The reality is if you take a PC and switch it on, data is being collected. There is data called telemetry. When you plug a new printer into your computer, you expect it to work. It works because there is data being collected about what the printer is and it is being sent to different places, and a small piece of software is installed to make the printer work seamlessly, without the user having to interfere at all. We have had to engineer the operating system so that it is fully compliant with GDPR to allow that kind of system to work, but it works very much in the background. We have built the controls that allow people to determine how that data is collected. The EU-US Privacy Shield ensures that all data is treated completely consistently as between Europe and the US, but we probably need to define what we mean by data collection or use.

The Chairman: You could distinguish between functional data and personal data, could you not?

Hugh Milward: I think you probably could.

Lord Gordon of Strathblane: Another idea that might help the average user is frequently things come up, new terms and conditions, and you are asked, “Do you agree?” A mobile phone is fairly small and most people just press “I agree”. Would it not be a good idea, and indeed in your interests, to have your terms and conditions approved or given a kitemark, as it were, by some industry body or co-regulatory body that would simply say, “You are not signing away your house if you sign this”?

Katie O'Donovan: All our terms and conditions have to be compliant with GDPR. It is not an industry body, but it is a legal standard that our terms and conditions have to meet. We have worked hard to make meaningful alternatives available to people too. On YouTube our community guidelines are written very succinctly with cartoon images to illustrate what is meant by those. On Family Link, which is our product for families, and YouTube Kids, we have on-boarding flows that are written in a language and style that is very succinct and easily digestible. We are working to ensure that people understand in a meaningful way how we engage with them.

Baroness Quin: Could the Cambridge Analytica scandal, which was very concerning, happen again, or do you feel that the systems now in place would prevent data being harvested in that way?

Rebecca Stimson: After that happened, there were extensive changes made to the platform as to how apps can and cannot engage with users’ data and the control that people have over them. We changed things such as the default settings for interacting with apps and so forth. I would be a bit too brave if I said that something like that could never happen again, but the way that happened has been addressed by all the changes we have made to the platform since. Certainly the evidence we see is that people understand that when they are interacting with social media they are exchanging data and data is being collected. In addition to the kinds of tools that we all have on our platforms, and the availability of our policies, to an awful lot of people it would not matter if they read them anyway; they would not understand. Thus the other half of that coin is about education and support for people to become digitally literate and savvy about what they are seeing. We give an enormous amount of information and transparency about data collection and what we do, but you have to meet the other half of that, to ensure people actually understand the full extent of what they are reading.

Lord Gordon of Strathblane: Most people realise there is a trade-off between you providing very good services, permanently, for nothing, and them providing information about themselves which is useful to you in terms of targeted advertising. What they might be less keen on is you selling that data on and becoming part of a data market.

Rebecca Stimson: We do not sell people’s data. That is not how our business model works. If you think about it logically, advertising is what underpins our business, so people’s data is extremely valuable to us, and it would make no sense for us to sell it on. We do not sell people’s data.

Lord Gordon of Strathblane: Could I put a point to Katie? There is a moral difference between looking at what I use Google for, searching for something, whether it is Wellington boots or a holiday in Athens, and targeting advertising at me, which is probably useful to me as well as useful to you financially, and scanning my emails, which you used to do, to see if there was anything. You gave that up. Why did you give it up?

Katie O'Donovan: I do not know why we stopped doing that. I think the intent behind it was the same as showing you adverts for Athens, in that we felt that we may be able to provide some services which were of utility and relevance to the email, but we have stopped doing that.

Lord Gordon of Strathblane: How do you monetise Gmail?

Katie O'Donovan: Some of our products are free to use and do not carry advertising. Some carry advertising on relevant searches. We operate Gmail under the umbrella of Google.

Lord Gordon of Strathblane: If somebody said to you that Google should be split up because it is too large and that Gmail should be a stand-alone service, it would have to close?

Katie O'Donovan: I would not want to hypothesise as to what would happen in those circumstances.

The Chairman: But in itself it would not be a successful business model?

Katie O'Donovan: Again, I have not looked closely under the bonnet of Gmail, but certainly at the moment I believe that it would require a different business model to sustain it.

The Chairman: There is a final point from Baroness Kidron and then we will close.

Baroness Kidron: It has come up a couple of times, the fact you do not sell people’s data, but is it fair to say what you are doing is selling the user to the advertiser? In that exchange, given your share price and the bank accounts of most users, that is quite valuable. The users’ attention is quite valuable to you, and that is why you have designs that encourage use? Would you say that is a fair analysis?

Rebecca Stimson: As I have said, we have re-engineered how Facebook works that has delivered less use, very specifically. When it comes to advertising, we have certainly found when we have done surveys into this that people understand the deal: we have data on them and the advertising means that our service is free for them to use at that point. When you ask them if they would rather have relevant or irrelevant advertising, they say, “If we have to have it to have a free service, we would rather that advertising was relevant”. That is the exchange and trade-off that happens between us gathering people’s data and targeting useful advertising at them, which most people find helpful, rather than the platform being paid for by advertising that is completely irrelevant to them.

Katie O'Donovan: I do not think that is a fair or accurate reflection. Our business model is different from Facebook’s in this instance. We run adverts on a small proportion of searches which are relevant to those search terms, where an advertiser will pay when someone clicks on that link. It is not about keeping people on a site or using them as a commodity. It is about relevance and helping people find the information that they need on both a commercial and non-commercial basis.

Hugh Milward: We have a number of different business models that have different monetisation plans around them. Our search engine is very much the same as Katie has mentioned. We have our mail service—originally called Hotmail—and that is self-sufficient in its own right. It is now called Outlook. That is funded by advertising but it is break even in terms of cost.

The Chairman: May I thank our witnesses for their evidence? I am sure you think we have asked quite enough questions, but it may be there are areas that we have not touched on that you would like to comment on. We are going to ask you to offer some clarification in writing on a few points that we discussed earlier and, at the same time, anything you think might be useful to the Committee would be welcome. Do our witnesses have anything they would like to briefly add at this point?

Katie O'Donovan: Thank you very much for inviting us to give evidence.

The Chairman: Thank you for coming and answering our questions. As I say, the clerk of the Committee will be in touch to follow up on a few points.