Carnegie UK Trust written evidence (DAD0087)


  1. Carnegie UK Trust was set up in 1913 by Scottish-American philanthropist Andrew Carnegie to improve the well-being of the people of the United Kingdom and Ireland, a mission it continues to this day. Carnegie particularly charged the trustees to stay up to date and the trust has worked on digital policy issues for some years.


  1. In early 2018, Professor Lorna Woods (Professor of Internet Law at the University of Essex) and former civil servant William Perrin started work to develop a model to reduce online harms through a statutory duty of care, enforced by a regulator; the proposals were published in a series of blogs and publications for Carnegie and developed further in evidence to Parliamentary Committees[1]. The Lords Communications Committee[2] and the Commons Science and Technology Committee[3] both endorsed the Carnegie model. In April 2019 the government’s Online Harms White Paper[4] proposed a statutory duty of care enforced by a regulator in a variant of the Carnegie model. France[5], and apparently the European Commission, are now considering duty of care models for online harms.


  1. CUKT are producing a draft bill to implement such a regime, based upon our policy document of April[6]. We shall publish the text of a draft bill shortly after the election.  If it is acceptable to the Committee, we shall submit this to the Clerk as supplementary evidence. As the bill is not yet ready, this paper explains how the teams thinking would apply to harms to the democratic process. We set some of this out earlier in the year in a joint statement with other organisations.[7]



Code is law


  1. Online platform services take their shape from design decisions taken by people who run them. Everything that happens on an online platform service is a result of decisions taken by the company or person that operates the platform. These include decisions about terms of services, decisions about software, decisions about the resources put into enforcing the terms of service and maintaining the software.[8] These decisions are not neutral in terms of content. They create tools that nudge users to behave in certain ways; reward certain types of content and its sharing or spread; reinforce certain behaviours and promote and suppress content. As internet enthusiasts say, within a software environment ‘code is law’.
  2. Those that control platforms make decisions that enable or otherwise: targeted advertising and the rules surrounding it, a highly personalised view of data, algorithmic acceleration of dissemination of some content, some forms of speech but not others, the tools available to users to shield themselves, transparency about the platform operation, the speed of commenting, speed of sign up etc Harms to the electoral system that the Committee might identify could arise from these decisions.


Regulatory system


  1. A regulatory system could influence the companies’ decisions and reduce harms at a systemic levelBased on an analysis of existing regulatory models, Carnegie UK Trust (CUKT) proposes that platform operators should be obliged to take reasonably practicable steps to prevent reasonably foreseeable harm arising as a result of the operation of their platforms. Harms would include harms to the democratic process, whether through manipulation of voters through the spread of ‘fake news’ or microtargeting of political ads; or because of the silencing of minorities through online abuse and harassment (as noted by the Joint Committee on Human Rights in its recent report.[9])


  1. The operation of this statutory duty of care regime should be enforced by OFCOM, working with sector-specific regulators such as the Electoral Commission and the Information Commissioner. OFCOM is not perfect, but it is available to start work quickly and has a strong track record of evidence-based decision-making, standing up to global companies and experience of dealing with issues that involved balancing human rights. A new regulator would have none of these attributes.


Operation of regime in respect of harms to the democratic process


  1. In our draft bill which we shall submit to you shortly, we set out a duty of care, define what is covered by the duty and the high level, systemic actions online platform service operators would have to take to deliver the duty. The draft bill amends OFCOM’s duties to make it the regulator and gives them the power to implement the regime, including to draft high-level systemic codes of practice.  We specify a very small number of specific harms that companies must aim to reduce – harm to the democratic process is one of these, alongside national security and child sexual exploitation and abuse (CSEA).


  1. The regime we set out is generic, like the Health and Safety at Work Act 1974 regime.  Defining the regime at a generic level ensures it will be long-lived and systemic in effect. Many of the design features of services which give rise to a risk of harm do so in relation to a range of different harms.  In the Carnegie regime, we propose that OFCOM approves codes of practice on each of the following (non-exclusive) generic factors:


  1. If potential harms to the democratic process are fed into OFCOM’s generic code-making process then they can be engaged by the online harm regime to some extent. Codes can be advisory or, after a parliamentary process, statutory. This then allows the Electoral Commission to focus on areas of detail not dealt with by systemic regulatory action.


  1. As an example, issues with malign creation of fake accounts or abuse of women and minorities have impact far more broadly than just the democratic process. They would be dealt with in the generic codes. Issues about imprints on online electoral adverts would be more appropriate for a specialised process.


  1. Online platform service operators themselves will have to consider harms to the democratic process in their risk assessments and mitigating actions. We would expect OFCOM to work with the Electoral Commission, political parties, online platforms and civil society to produce a code of practice.  OFCOM could even adopt a code made by the Electoral Commission. The code could outline high level harms in this area. 



  1. OFCOM’s enforcement process would work as follows. If OFCOM had reason to believe that harm was arising in respect of electoral processes in the operation of a platform service, OFCOM could make an information request addressed to the company concerned. If the results gave rise to concern, OFCOM could begin an assessment process. OFCOM could subsequently make enforcement and ultimately penalty notices.   In assessing penalties for any offences OFCOM would bear in mind the extent to which an operator has complied with codes, or any directions from Ofcom such as information notices.  OFCOM would have the power to fine or in extremis, require ISPs to block services.

Benefits to the electoral regulator of working with duty of care regulator


  1. OFCOM would be tasked to work with the Electoral Commission[10], which would give both regulators benefits. Working with OFCOM would offer an electoral regulator the following benefits:


Route to legislation


  1. Carnegie has produced a draft bill of about 60 clauses and a paving bill of four clauses requiring the government to get on and appoint a regulator to prepare for the new regime. Our intention is to move debate on to vital details. Government has been delayed by events and their turbulent wake.  We are concerned that if government proceeds at its current pace no regime will be in operation until 2023.


UK /USA trade deal – potential restrictions on implementing Committee recommendations


  1. The Committee, after its deliberation, might wish to make recommendations that require the online platform service companies to do things. We should flag up, as doubtless others have, two potential structural barriers implementing such recommendations that might arise from hurried trade deals with the USA and the EU.


  1. In the USA, tech companies are lobbying for the preservation of low levels of intermediary liability in overseas markets through trade deals. The current USMCA deal contains such previsions (S19.17)[11]. The USA Trade Representative has said that maintaining this is a priority in the UK/USA talks. We don’t know of course how this would be done but there is a chance it could have an effect.


  1. The forthcoming EU ‘Digital Services Act[12]’ will review the limited liability of the E-commerce directive. The Commission has reportedly written to France to ask that France delay implementation of their recently passed anti-hate speech law[13] and instead wait for the new Digital Services Act. France has said that they will not wait.  It is possible that, if the UK were to accept the current single market laws on communications, as industry wants, then a similar request would be made of the UK in respect of online harms to wait for the DSA but in a trade deal context where the ability to act or not has to be weighed against other interests.


Republic of Ireland


  1. We commend to the Committee the work underway in the Republic of Ireland to reduce harm to children and others from online media and to implement the Audio-Visual Media Services Directive[14]. Under the country of origin principle in broadcasting regulation, the Irish rules will govern YouTube and Facebook video in Europe. Where the Committee makes recommendations with regard to video online and aspects of video services, the Irish work will have a bearing. The Irish government’s proposals are due for publication in December/January. The Irish government also recently announced plans to bring in a new law to regulate online political advertising during elections.[15]




Carnegie UK Trust

December 2019



[1] Our work, including blogs, papers and submissions to Parliamentary Committees and consultations, can be found here:







[8] See for example work by the Behavioural Insight Team (April 2019), which looked at eg, how search results on mobiles influence our choices, how little we engage with online privacy notices and impacts on attention and working memory: The Behavioural Science of Online Harm and Manipulation; also work by 5 Rights (July 2018) looked at how design features like auto-play, auto-suggestion, Likes, re-tweets, notifications, buzzes, pings, typing bubbles, streaks etc shape childrens interactions online: Disrupted childhood: the cost of persuasive design; and Norwegian Consumers Council (June 2018) research on how the online design environment shapes consumer choice: Deceived by Design