LINX – written evidence (IRN0055)

“The Internet: to regulate or not to regulate?”

About LINX

1. The London Internet Exchange Ltd, “LINX”, is a membership association for network operators and service providers exchanging Internet traffic. It is part of our core mission to represent our members’ interests in matters of public policy.

2. Established in 1994, we have approaching twenty-five years’ experience engaging with policymakers and other external stakeholders in matters of Internet regulation, including regulation of content and end-user behaviour.

3. With more than 780 member organisations, including most major UK ISPs and most formerly-incumbent European operators, we believe we have highly informed expertise.

Introduction

1. The Internet is often characterised by policymakers as “the Wild West”. Certainly the demonization of Internet companies by political leaders, pundits and media often has the appearance of an angry mob bearing pitchforks and flaming brands. As in the Wild West, many political leaders often seem concerned only with drumming up a posse to ride out to deal rough justice to assumed wrongdoers in black hats. There is little place in this discourse for careful balancing of competing rights, the possibility that serious wrongdoing and minor or even merely unpopular infractions are being conflated, or the social value of a system of due process and fair treatment.

2. Policy makers, including even the Prime Minister, are fond of trotting out the mantra that the “the law that applies offline must also apply online”. However, in more than twenty years experience engaging in Internet regulatory policy, it is our experience that what is truly meant is that the restrictions and prohibitions that apply offline must also be enforced online – and that they must be enforced by private companies. We detect no corresponding eagerness to ensure that the administration of online justice is protected by the same traditions of the independent prosecutors, tribunals and courts, nor even that private companies attempt to uphold and replicate the similar principles that underpin those institutions. Instead, Internet companies are judged solely by the speed with which they removed material alleged to violate standards that are unclear and untested, and by whether campaigning activists or politicians can find on those companies services examples of material or behaviour that outrages the public conscience.

3. In a mature, civilise democracy, it is accepted that the administration of justice balances competing interests. The public interest does not only lie in protecting members of the public from malefactors, but also in providing the public also a clear, comprehensible and predictable standard of laws by which they can regulate their own conduct. The public interest is not served only by the swiftest enforcement action, but also by a fair process.

4. The longstanding aggressive focus of political debate on achieving more vigorous enforcement obscures the need to develop a balanced and stable framework for the resolution of disputes and the administration of just decision-making concerning law enforcement in the online environment.

5. LINX therefore welcomes the broad scope of this Committee’s enquiry, and hopes it will provide an opportunity to reflect on what kind of society we want to build in the Internet age, and what framework of responsibilities for the State and private companies would bring that about.

6. In this submission, we will identify some matters of principle the Committee may wish to consider, political leadership on which we consider essential to make material progress in developing a new framework.

Identification and characterisation of illicit content/behaviour

7. Policy discourse about Internet regulation is usually conducted in a piecemeal fashion, focussed on one type of illicit content or behaviour (e.g. terrorism or fake news). This committee’s enquiry is a welcome exception.

8. Unfortunately, when it comes to the application of policy to real-world instances, a discussion about illegal content jumps the gun: it presupposes that the topic is, for example, terrorist content, and demands action in response.

9. In real life, a business process within an Internet company never starts with “terrorist content”. It always starts with an item of content, and the first step is to determine whether it is indeed to be treated as terrorist content.

10. The first question faced by an actual Internet business is therefore never, “what do we do when we encounter content of type X?”, but “Does this item trigger the process we follow when we encounter content of type X?”. In other words, is this really X at all?

11. This remains fundamentally true even if the business accepts decisions from an outside source. You can, if you wish, outsource the decision-making on whether to apply a particular rule (such as deletion) to particular classes of content (such as terrorist content) to a third party. But you still need to decide which third parties, and what the scope of their authority will be (for example, that CTIRU will be allowed to designate material as terrorist, but not as contrary to our policy on abusive speech), whether a particular report did in fact come form that authority, and whether that report was made in accordance with the agreement or legal obligation to accept such reports.

12. There is no escaping the fundamental question. You can have a rule requiring suppressing X, but is this item before me X?

13. For this reason, political discourse on Internet regulation would be greatly improved if all agreed to cease to frame the question as “what should be done about illegal/illicit/harmful content”, and instead frame it as “what should be done about reports (or allegations) of illegal/illicit/harmful content?”.

14. In other words, when making recommendations for dealing with illegal, illicit or harmful content, there needs to be recognition that whether a particular item infringes is not intrinsically known. The decision-maker, the process by which they reach a decision and the standards (including scope of rules) the decision-maker applies, all need to be identified. If these aspects are not clear, or do not stand up to scrutiny, then the recommendations are unlikely to be a useful contribution to the development of Internet regulation. And if they depart materially from the norms and practices that apply offline, then these variations should be reasoned and justified, and the pretence that they “merely implement the same law that applies offline” should be dropped.

The significance of an intermediated society

15. In many respects, the Internet is no different from ordinary, offline society. Most of the harms so frequently cited as being facilitated by the Internet are merely examples of ordinary criminal behaviour and the frailty of the human condition: the Internet’s only contribution being that as well as facilitating and empowering the good, like any bare tool it also can also be used for evil.

16. There is, however, one material distinction between ordinary offline society and the Internet: on the Internet, almost any action or behaviour is wholly reliant on one or more intermediaries: the Internet access provider, the domain name registry, the web site or social media platform, the search engine etc. In the offline world, an actor can frequently act on their own, and are alone accountable for their action. In the online world, if a necessary intermediary chooses to intervene to suppress the action, the actor is sanctioned.

17. The legal-social model in the offline world, at least in the United Kingdom, presumes freedom under the law. An actor is capable of acting freely, and unless the law prohibits a certain action he will not face “law-enforcement” constraints. A member of a free society practically enjoys the freedom that the law allows.

18. By contrast, in the online world, when the Internet industry as a whole is forced by political pressure to intervene to suppress not only that which the law prohibits, but also that which political leaders state “any responsible company must act against”, the ordinary public no longer enjoy their full measure of legally permitted freedom. In a sense, the law has changed, to become more restrictive.

19. In practical effect, online freedom is no longer protected by the principle of freedom within the law, but rather by the practical difficulties Internet intermediaries have in complying with and enforcing such political directions.

20. Maybe greater restriction is desirable. In some cases, this is very probably true. But if the democratic organs of society took responsibility for those changes, their working methods and tools (legislation) would provide ancillary benefits of public interest: fore notice, specific rules with codified and knowable scope and limits, the opportunity to challenge compatibility with human rights norms, and more. When such changes are introduced through changes in corporate policy, these benefits can be hard to realise. And that is quite apart from the intrinsic value that lies in having enforceable rules determined democratically.

Legal exceptions and justifications

21. If our aim is to ensure that the law applies online as it does offline, that means applying not only the headline prohibitions, but also the limitations of scope, exceptions and legal defences to those same prohibitions. Failure to do so would result in the scope of the prohibitions being effectively extended in the online environment or, to put it another way, freedoms being restricted that Parliament sought to protect.

22. Political pressure on on private companies demands rapid enforcement action, and usually deems enforcement of the law a minimum requirement; indeed, much political discourse assumes that merely upholding statutory prohibitions falls below the minimum requirement, and that any responsible Internet company ought to go much further in the “protection of the public”, or lose its license to operate.

23. We do not deny that companies will often want to place restrictions on the use of their service that go well beyond legal requirements. It is perfectly legitimate for companies providing a shared communal space for their user community to want to regulate that so that it provides a welcoming and enjoyable user experience. We certainly do not want to suggest that any member of the public has the right to use any Internet-based service in any lawful manner, free from any form of restriction or limitation by the operator of that service.

24. At the same time, we must recognise that when the Internet industry as a whole responds to political pressure to be much more restrictive than the law requires, this effects a qualitative change in the impact of the law in question, as a tool for regulating social behaviour.

25. We will draw on two examples to illustrate this point, one drawn from the criminal law, and one from the civil. In both cases, the examples relative to activities/infringements that can take place wholly online.

26. In the criminal sphere, Parliament establishes precise boundaries to prohibited expressive conduct. Not only does this better guide the public and allow people to regulate their behaviour, it also carefully balances competing interests. Occasionally, such bounds are matters of great political controversy, and for private companies to ignore them (or for law enforcement and companies to collude in so doing) would be to give great disrespect the democratic process.

1. For example, in the debates in Parliament about the extension of the law prohibiting incitement to racial hatred so as to also criminalise incitement to religious hatred, there was passionate discussion about whether this would unduly chill freedom of speech, both in the form of proselytizing and of legitimate criticism of religious dogma. Ministers answered that the Bill did not prohibit incitement of hatred against a religion (which is an idea), but against people defined by their adherence to a religion – and said that this is a crucial distinction. Moreover, Ministers argued, no prosecution can proceed without the consent of the Attorney-General, which is a powerful procedural tool for ensuring that the new law did not become a blasphemy law by the back door.

2. In the Internet context, however, politicians regularly demand that Internet companies act against “hate speech”, an undefined category that includes, but is certainly not limited to, incitement to religious hatred as defined in English law. Many companies prohibit “hate speech”, with no further definition, in their terms of service. And the reports they receive, demanding suppression of instances of alleged violations, are certainly not limited to those that have been approved by the Attorney General.

3. What price, then, Parliament’s passionate debates and careful compromises? Are these of no lasting importance? Is it that Parliament did its job by creating a law, but the baton is now passed to companies to take it much further? Are the prohibitions that Parliament enacts of vital importance to apply to Internet communications, but the exceptions and limitations of scope that Parliament also enacted merely a trivial matter easily dispensed with? And if not, how are companies supposed to apply such limits? Are they supposed to step into the shoes of the Attorney General, and second-guess what would be seen as proportionate? For that matter, are companies supposed to tolerate heated and intemperate attacks on a religion, as long as they are clearly targeted only to that religion’s ideology and dogma rather than its adherents? When are companies free to go beyond the law, and when ought they to seek to replicate its effects? These are not easy questions, but the answer “Companies must prohibit their users from doing or saying anything Parliament has prohibited, but need not pay any attention to the limits to those prohibitions or the freedoms Parliament has protected” seems unsatisfying as well.

27. For an example from civil law, consider the regulation of the use of trademarks in domain names.

1. In the context of trademark infringements, it is usually an infringement to use a registered trademark in the course of business without authorisation. However, while not everybody immediately recognises that this is an unduly simplistic description of trademark law, nobody would really think that this prohibits a ratings service like TripAdvisor (or, indeed, longer-established variants on the same theme, like Which?) from referring to trademarked brands in critical reports, notwithstanding that TripAdvisor and Which? are both doing so in the course of business.

2. The Internet, though, regularly throws up further challenging examples. DNS domain names are an essential resource for anyone seeking to offer services online. Registration of a domain name is entirely dependent on a particular kind of Internet intermediary, the domain name registry.

3. Domain name registries quickly encountered the following question: is it acceptable for a disgruntled customer to register the domain EXAMPLEBRANDNAME.COM (where “EXAMPLEBRANDNAME” is a registered trademark), for the purpose of running a website disparaging the company that trades under that trademark? What about “EXAMPLEBRANDNAMESUCKS.COM” ?).

4. Some domain name registries have developed their own system of private law that attempts to approximate, but not replicate, trademark law. Nominet (which operates the domain registry for domains ending in .co.uk) has its Dispute Resolution System. ICANN, which sets the policy for a wide range of domain registries, including the registry for .com, has its Universal Dispute Resolution System. Each applies its own systems of standards and sanctions.

5. Each of these dispute resolution systems has been criticised as being overly favourable to trademark owners, both by being overly restrictive of unregistered use (i.e. taking action to suppress unregistered use that a trademark court would not act to suppress) and by being excessively punitive (i.e. following a trademark owner’s complaint about misuse of a domain, being willing to transfer the domain name into the ownership of a trademark holder in circumstances where a court might award a less draconian remedy, such as a small monetary award or even an undertaking to refrain from further infringement).

6. Ought DNS Registries to seek to replicate, rather than merely approximate, trademark law? If a domain registrant can show that a court would not seize his domain and transfer it to a trademark owner, should a Registry exercise similar restraint? Or do only some parts of the law apply online?

Evidence and context

28. The previous sections showed that there is often no clarity as to whether an Internet intermediary should seek to replicate the effects of the law as it applies offline, in all its complexity and careful balances of legitimate competing interests, or whether companies should simply aim to suppress the mischief such laws aim to prevent, without too much care for the interests of the accused.

29. In this section, we shall argue that even when there is agreement as to the meaning of the law, an Internet business is frequently incapable of applying it in a manner even broadly equivalent to the way it would be treated by a court.

30. When a court – or even a law enforcement agency, or some other administrative body or tribunal – considers an allegation, they take account of the evidence. The context in which that evidence occurred may well be relevant.

31. When an Internet intermediary is faced with a report alleging wrongdoing, that report is more closely analogous to an indictment than to evidence. It is, at best, one presentation of some of the facts. The business has no ability to call for more evidence, and little capacity to question what is there (even when so minded). In many cases context will be entirely missing, and the context will very often be sparse and misleading.

1. For example, social media platforms are often called upon to adjudicate accusations of harassment. In so doing, they will have before them one message, or perhaps one series of messages, between the parties. In most cases, they will have little idea of previous context and no ability to discover further background – and no capacity to adjudicate it if they did. In this context, harassment (which properly describes a sustained and abusive pattern of behaviour) quickly devolves into “he/she called him/her a mean name”. When dealing with common patterns, such as bullying between school-age children or students, the likelihood that the complainant is truly the victim rather than the perpetrator gaming the social media platform’s systems to inflict further pain is not a great deal higher than even.

2. This is not to say that social media companies should not have rules against harassment, or that they should not act on complaints of abuse. It is, however, to say that expecting to rely on social media platforms as a solution to the phenomenon of bullying or adequate protection for victims of harassment is profoundly mistaken. On the contrary, even to describe them as “having an important role to play” creates quite unrealistic expectations about how such companies can adjudicate disputes or meliorate social problems. It would be better to recognise that the extent of their capability barely even extends to being able to protect the integrity of the environment of their own service.

32. To give a very different example, consider the phenomenon of “fake news”.

1. Since free societies such as the United Kingdom do not actually criminalise the publication of statements, opinions and characterisations of news events that someone believes to be untrue or misleading, perhaps the closest legal analogy is defamation. A court hearing regarding defamation will not be limited to consideration of the article that is the subject of the case; there will also be evidence about it, and about the parties.

2. Social media platforms are under intense political pressure to “combat” so-called “fake news”. Let us set aside questions as to what would really constitute fake news (the committee can consider that as belong to the section immediately above, which could probably be expanded to a book-length treatment were we to attempt to deal with that category comprehensively), and also setting aside whether the real concern is State-on-State psychological warfare (and whether that is not more properly a matter for national counter-espionage agencies than multinational social media websites), and consider only the verification of news. What would social media sites need to do to be able to make even a credible effort at verifying whether a particular news item was correct? Would they need to be able to compel evidence, like a court? Would they need to have on-site reporters, who took evidence from multiple sources, like a news agency? Would they have to go further than merely marking something as “approved” or “fake”, into publishing their own opinion so as to get the requisite level of nuance?

33. In this context, it’s worth noting that the New York Times ran an editorial on 21st April 2018 discussing fake news on Facebook, in which it proffered the headline “Palestinians Pay $400 million Pensions For Terrorist Families” as an example of “far-right conspiracy programming” demonstrating that Facebook was rife with fake news. The NYT subsequently retracted this[1], after Jewish magazine Tablet, amongst many other, pointed it to the well-documented financial support programme the Palestinian Authority runs for the families of persons incarcerated or killed by the state of Israel[2]. 

Proportionality, total enforcement and law enforcement by algorithm

34. Once a decision has been made that a particular item matches the description given to it, and is illegal, what should happen? Should all content that is illegal be suppressed?

35. In the “offline” world, total enforcement is not the usual model. On the contrary, a great deal of discretion is available to law enforcement officers and to prosecutors about whether to take any enforcement action at all against law-breaking, and if so, whether to pursue coercive sanctions or just have “a quiet word” with the miscreant.

36. The desirability or otherwise of total enforcement is one of the (relatively few) areas where the Internet’s technical features really are relevant to the policy options.

1. Prior to the Internet, the most important forms of expressive communications that most people would engage in would be private conversations – in the home, in the workplace, in the pub. In each case, people spoke to a limited audience.

2. It was neither feasible nor proportionate to attempt total enforcement of laws prohibiting certain types of expressive communications in that context; actual enforcement in such a context would be reserved for the most serious cases, such as where an abusive shouting match turned into violent affray.

3. By contrast, a relatively small number of people had access to wide audiences, through broadcasting, publishing of newspapers and books etc. Compliance with the law was treated much more seriously by their publishers than private communications. The number of people who felt the bite of such restrictions was much smaller, and it most commonly felt in a professional context rather than their personal life (and so came with an expectation of a professional level of behaviour).

4. With the advent of the Internet, “online” is added to “at home”, “at work” and “in the pub”. Indeed, “online” partially replaces those previous contexts: a great deal of the communications people would previously have done only face-to-face in those environments now takes place in the moment, online.

5. That audience has changed somewhat too – ordinary members of the public are now capable of reaching large numbers of people for the first time. They also sometimes achieve that unintentionally or carelessly.

6. Even for public figures and media professionals, the advent of social media has dramatically changed the context of their writing. Whereas previously a press release, TV interview or opinion column would be a carefully considered (and likely planned and vetted) statement, now 140 characters are tapped out on a phone in an instant, and sent off into the world to face perhaps widespread outrage.

37. The world is therefore wrestling with the question of how to cope with the public audiences being available to the untutored general public. Should they be held to the standards of traditional publishing? More particularly, should they be held to the standards of law that were previously usually only applied in practice to traditional publishing, even if they theoretically covered private communications? It would be ironic if the advent of the Internet led to such a marked diminution of freedom of expression.

1. It seems unlikely that the answer to this can be as simplistic as “the same laws must apply online as offline”; for most people, in most contexts, those laws never did bite on them offline for any practical purposes.

2. At the same time, abandoning all legal regulation of expressive communications is also utterly out of the question.   
2. Perhaps the answer lies in recalling the broad discretion that prosecutors and law enforcement have always had to turn a blind eye to illegal expressive conduct.

1. Should Internet companies have such discretion? If so, how should they exercise it? What guidance should they have on when it is proportionate to intervene, and when not. And do they have a duty to disapply the law even-handedly?

2. We may welcome prosecutorial leniency towards unwise but ultimately minor expressive lawbreaking: not every fib should be prosecuted as fraud, nor every tirade prosecuted as a cause of needless distress.

3. But if we felt that prosecutorial discretion was being applied with systematic bias, we would feel aggrieved[3]. Internet intermediary companies are currently under no legal duty to conduct their administration of laws in an even-handed, politically-neutral fashion.

39. One option would be legal reform: the Law Commission might review the breadth of laws that can be infringed, perhaps inadvertently, online, and analyse where it was previously assumed that judgement would be exercised in their enforcement, and make recommendations for change so that these laws could be administered not by public institutions with duties to the rule of law, including judgement as to reasonableness, but by corporations with no such duties, and algorithms with no such judgement. This does not seem an easy task.

40. Another approach would be to decide that placing such great weight on private companies to administer the law is a mistake. Instead of reforming the statute book, the court system might be modernised to make it capable of administering justice for an online society. Specialist courts and tribunals are rightly not created easily, but they are created when a pressing social need for them appears: we have the Patents Court, the Small Claims Court and the Employment Tribunal. Why not an Internet Abuse Court, where cases of wrongdoing are managed through a streamlined online process, using public APIs that enable Internet companies to refer suspected criminality for action?

41. We do not pretend to have an easy answer to these problems, but the current situation is untenable. Internet companies are already struggling with the volume of complaints and allegations, and political leaders are quite insistent that what happens today is inadequate. The largest companies are being pressured into greater investment in, and use of Artificial Intelligence – algorithmic decision-making. However, far from a magic solution, AI raises serious concerns of its own.

42. The issue of “filtering”, or the algorithmic prior-restraint suppression of Internet content has special relevance to concepts of proportionality and total enforcement. An algorithm knows no compromise, understands context barely if at all, and applies its rules ruthlessly. If an article satisfies its pattern-matching rules, the filter comes down in force: it has no ability to say “this is only a minor breach, and in the context of a broader and more acceptable pattern of communication, let it pass”. The computer will say No.

43. There is much that can and should be said about algorithmic enforcement in relation to how crude it is, and the manifold gross opportunities for over-blocking. Such opportunities include programming mistake (known in the industry as “the Scunthorpe problem”, after primitive filtering systems that were intended to block profanities), the impossibility of fully capturing the nuances of law and social norms within the precise but unreasoning language of computer code, and the algorithm’s inability to understand and take account of crucial factors such as context and intent. We assume the Committee will hear from many other witnesses on this topic. Instead, we leave the Committee with a more philosophical question: even if our robotic censorship engines had perfect accuracy in restraining criminal speech before it ever took place, would you really want a society in which there was 100% machine enforcement, all the time, without proportionality, tolerance or mercy?

44. We caution the Committee not to dismiss this question on the grounds that we are far from facing that problem. Algorithmic enforcement (filtering) is already in place in some companies, searching for some kinds of content. It is certainly far from 100% accurate, but equally it is certainly 100% enforcement, in that any and every article that matches the algorithms’ pattern-matching tests is marked and treated accordingly, as “X” type of content[4].

Corporative incentives and liability

45. Courts are immune from liability for their judgements. Other tribunals can and do have limited or broad immunity. Internet businesses do not.

46. On the contrary, while Internet “hosting providers” have protection from liability for content stored on them by their users until they have “actual knowledge” of that content, that protection evaporates when the company acquires actual knowledge. From that point forward, a company risks being sued alongside the user; as an entity that is easily found and has deep pockets, it may well be the primary target of such a lawsuit. This provides a further reason why an Internet company is ill-suited to administer questions of law and justice; it is not independent and even a good faith attempt to reach a fair and just decision could land it with steep financial liability. The incentive is to suppress whatever provokes a complaint, regardless of the merits, and when Internet businesses are more protective of their users than not-at-all, that is to their credit.

Answers to questions

Question 1: Is Internet-specific regulation needed

47. For the most part, the mischiefs that regulation seeks to correct are not specific to the Internet. Most of the problems people identify with content and behaviour are problems that exist in the offline context, and for which viable laws already exist. New regulation should only be introduced where there is a proven need, and a realistic prospect that the regulation would improve the situation: we do not favour the introduction of legislation merely to “send a message”.

48. One area we do identify a problem is in the biased incentives Internet intermediaries face in the context of Article 14 of the E-Commerce Directive: hosting providers only have protection from liability until they have actual knowledge of the article. This places on them a strong incentive to remove material, for fear of taking liability alongside the publisher, inhibiting their ability to make a fair and independent assessment of the validity of any complaint.

Question 2: liability protection for platforms

49. Protection from legal liability for third party content is an essential part of offering an intermediary service. Without such protection, it is impossible to risk carrying material that the company itself does not consider legally safe; effectively, endorse.

50. If a platform is held liable for third party content, the platform can only permit such content as it approves. This effectively extinguishes its role as a platform for the expression of a diverse range of views and content, and reduces it to a single-publication, akin to a magazine.

51. Accordingly, liability should only exist for content that the platform creates, selects, or promotes with endorsement.  

Question 3: Platform moderation

52. Platforms should aspire to a high level of transparency in setting out their terms of service and acceptable use policies, so that their users can know how to regulate their own conduct in conformance to those policies.

53. The largest platforms, those with a global user base and which have the largest social impact, should consider the value of voluntarily foreswearing the “freedom of contract” right unfettered and arbitrary choice as to who may use their platform, and instead hold that it is available for use by all, subject to the terms of service, and pledge to treat all users fairly in determining whether there has been a breach of those terms.

54. The specifics of what constitutes “fair treatment” of users will vary according to both the nature of the service and the nature of the moderation, but in general it refers to some form of due process/natural justice guarantees.

55. A range of specific measures could be considered by platforms, including

1. Notifying their users if a complaint is made against them, or otherwise their account or content is suspected of non-compliance with terms of service/acceptable use policy.

2. Informing any user found in breach which element of the terms of service they were found to have breached, and which action or content was found to have caused the breach, with sufficient specificity that the user can understand why they are being sanctioned.

3. Granting users the right to offer a defence that will be taken into account before they are sanctioned (e.g. before their account is closed and they are banned from the platform)

4. Granting users a right of appeal

5. Ensuring that decision on breaches are taken by an independent person

6. Taking measures to ensure that breach decisions are taken consistently and without bias, so that similarly situated users will be treated equally, regardless of viewpoint. The first step to doing this would be to adopt this as an objective, and to state that it was not the role of abuse management teams to promote the company’s own viewpoint or values.

56. It should be noted that offering procedural “due process” protections to users in no way fetters the platform’s right to determine what content is permitted on their platform, only how it acts in response to a suspicion or allegation that the rules have been broken.

57. Not all of these measures will be appropriate in all cases. For example, stronger protections might be offered when contemplated closing an individual’s (or a business’) entire account on a social media platform, than would be appropriate when merely removing one ephemeral communication from display.

58. Indeed, when removing certain forms of unlawful content, it will not always be appropriate to give the user advance notice.

59. Accordingly, the range of measures we suggest for consideration are intended both as a menu from which to select, and as a loose characterisation of measures that would need to be tailored to the platform’s circumstances.

Question 4: Role for the user community

60. User communities can sometimes play a useful role.

1. On large platforms, users can act as the service providers’ “eyes and ears”, identifying potential breaches of acceptable use for the service provider to review.

2. In some cases, where the platform is divided into chat rooms or forums that each have a long-standing community and traditions of their own, selected members of the community can be invested with supervisory powers.

3. Indeed, in the case of Wikipedia, the entire enterprise is largely made up of a self-governing community, with supervision from the formal organisation geared mainly towards process and mechanisms, and intervention of last resort.

61. In our view, the circumstances of different platforms, their services and their communities are so diverse that it is not useful to generalise.

Question 5: Protection for freedom expression and freedom of information

62. In our view, protection for freedom of expression and freedom of information is achieved in the first instance primarily by the wide diversity of Internet services that are available. Additional measures are needed only to correct a problem that this diversity does not adequately resolve.

63. We can identify two areas where freedom of expression and freedom of information may not be fully protected by the diversity of platforms available:

1. In the case of a very small number of platforms, the platform is so large and success, that a person’s freedom of expression might credibly be said to not be fully realised if they are not able to access a particular platform, because that is where the audience is.

2. When legislation, court orders, industry collaboration or government and other political pressure intervenes to ensure that the Internet industry generally follows one set of norms, then a competitive range of platforms does not suffice: the aggrieved user cannot simply go to another service provider because they must all act similarly in the relevant respect.

64. In our view, protections for freedom of expression within a particular platform are best achieved by commitments to transparency and lack of bias (discrimination either on the grounds of immutable characteristics or viewpoint), together with the possible adoption of certain due process mechanisms when considering suspicions of abuse/unacceptable content or behaviour.

65. We set out some suggestions for consideration in response to question 3.

Question 6: Personal data

66. GDPR introduces very substantial and indeed onerous new requirements in relation to the use that may be made of users’ personal data, and how users should be informed about those uses.

67. We do not think that any further requirements should be contemplated at the present time; better first to wait and see the effect GDPR proves to have in practice.

Question 7-8

68. No answer.

Question 9: Impact of leaving the EU

69. Article 15 of the E-Commerce Directive, which states that Internet intermediaries may not be placed under a “general duty to monitor” has never been properly transposed into UK domestic law.

1. Successive UK governments have taken the position that this was not necessary, on the grounds that no UK law does place intermediaries under such a monitoring obligation.

2. Nonetheless, the lack of such transposition leaves UK operators exposed to the risk that law may be interpreted to allow the imposition of such a duty. This is particularly severe in relation to laws that grant courts a broad discretion to impose poorly identified duties on third parties, e.g. s94A of the Copyright, Design and Patents Act.

3. While the UK was a member of the European Union, UK providers had the comfort that even though Article 15 had not been transposed, UK courts were still under a duty to act in compliance with EU law.

4. When the UK leaves the EU, this comfort is diluted (or removed, depending on transition provisions).

70. The protection from a duty to monitor is part of the core acquis underlying Internet regulation in the UK and EU. We recommend that the government proceed to transpose Article 15, with prospective effect, as part of the preparations for leaving the EU.

17 May 2018