Sunday 24 May 2020

A Tale of Two Committees

Two Commons Committees –the Home Affairs Committee and the Digital, Culture, Media and Sport Committee – have recently held evidence sessions with government Ministers discussing, among other things, the government’s proposed Online Harms legislation. These sessions proved to be as revealing, if not more so, about the government’s intentions as its February 2020 Initial Response to the White Paper.

As a result on some topics we know more than we did, but the picture is still incomplete. Some new issues have surfaced. Other areas have become less clear than they were previously.

Above all, nothing is set in stone. The Initial Response was said to be indicative of a direction of travel and to form an iterative part of a process of policy development. The destination has yet to be reached – if, that is, the government ever gets there at all. It may yet hit a road block somewhere along the way, veer off into a ditch, or perhaps undergo a Damascene conversion should it finally realise the unwisdom of creating a latter-day Lord Chamberlain for the internet. Or the road may eventually peter out into nothingness. At present, however, the government is pressing ahead with its legislative intentions.

I’m going to be selective about my choice of topics, in the main returning to some of the key existing questions and concerns about the Online Harms proposals, with a sprinkling of new issues added for good measure. Much more ground than this was covered in the two sessions.

Borrowing from the old parlour game, each topic starts with what the White Paper said; followed by what the Initial Response said; then what the Ministers said; and lastly, the Consequence. The Ministers are Oliver Dowden MP (Secretary of State for Digital, Culture, Media and Sport); Caroline Dinenage MP (Minister for Digital and Culture) and Baroness Williams (Lords Minister, Home Office).  

Sometimes the government’s Initial Response to Consultation recorded consultation submissions, but came to no conclusion on the topic. In those instances the Initial Response is categorised as saying ‘Nothing’. Some repetitive statements have been pruned.

Since this is a long read, here is a list of the selected topics:

1. Will Parliament or the regulator decide what “harm” means?

The White Paper said:

“… government action to tackle online content or activity that harms individual users, particularly children, or threatens our way of life in the UK, either by undermining national security, or by reducing trust and undermining our shared rights, responsibilities and opportunities to foster integration.”

“This list [Table 1, Online harms in scope] is, by design, neither exhaustive nor fixed. A static list could prevent swift regulatory action to address new forms of online harm, new technologies, content and new online activities.”

The Initial Response said:


The Ministers said:

Oliver Dowden: “The only point that I have tried to make is that I am just keen on this proportionality point because it is often the case that regulation that starts out with the best of intentions can, in its interpretation if you do not get it right, have a life of its own. It starts to get interpreted in a way that Parliament did not intend it to be in the first place. I am just keen to make sure we put those kinds of hard walls around it so that the regime is flexible but that in its interpretation it cannot go beyond the intent that we set out in the first place in the broad principles.” (emphasis added)

Caroline Dinenage: “For what you might call the “legal but harmful” harms, we are not setting out to name them in the legislation. That is for the simple reason that technology moves on at such a rapid pace that it is very likely that we would end up excluding something….  We want to make sure that this piece of legislation will be agile and able to respond to harms as they emerge. The legislation will make that clearer, but it will be for the regulator to outline what the harms are and to do that in partnership with the platforms.” (Q.554) (emphasis added)

The Consequence: It is difficult to reconcile the desire of the Secretary of State to erect “hard walls”, in order to avoid unintended consequences, with the government’s apparent determination to leave the notion of harm undefined, delegating to the regulator the task of deciding what counts as harmful. This kind of approach has serious implications for the rule of law.

Left undelineated, the concept of harm is infinitely malleable. The Home Office Minister Baroness Williams suggested in the Committee session that 5G disinformation could be divided into “harmless conspiracy theories” and “that which actually leads to attacks on engineers”, as well as a far-right element. One Committee member (Ruth Edwards M.P.) responded that she did not think that any element of the conspiracy theory could be categorised as ‘harmless’, because “it is threatening public confidence in the 5G roll-out” — a proposition with which the DCMS Minister Caroline Dinenage agreed.

Harm is thus equated with people changing their opinion about a telecommunications project. This unbounded sense of harm is on a level with the notorious “confusing our understanding of what is happening in the wider world” phraseology of the White Paper.  

Statements such as the concluding peroration by Baroness Williams: “I, too, want to make the internet a safer place for my children, and exclude those who seek to do society harm” have to be viewed against the backdrop of an essentially unconstrained meaning of harm.

When harm can be interpreted so broadly, the government is playing with fire. But it is we  not the government, the regulator or the tech companies  who stand to get our fingers burnt.

2. The regulator’s remit: substance, process or both?

The White Paper said:

“In particular, companies will be required to ensure that they have effective and proportionate processes and governance in place to reduce the risk of illegal and harmful activity on their platforms, as well as to take appropriate and proportionate action when issues arise. The new regulatory regime will also ensure effective oversight of the take-down of illegal content, and will introduce specific monitoring requirements for tightly defined categories of illegal content.” (6.16)

The Initial Response said:

“The approach will be proportionate and risk-based with the duty of care designed to ensure companies have appropriate systems and processes in place to improve the safety of their users.”

“The focus on robust processes and systems rather than individual pieces of content means it will remain effective even as new harms emerge. It will also ensure that service providers develop, clearly communicate and enforce their own thresholds for harmful but legal content.

“The kind of processes the codes of practice will focus on are systems, procedures, technologies and investment, including in staffing, training and support of human moderators.”

“As such, the codes of practice will contain guidance on, for example, what steps companies should take to ensure products and services are safe by design or deliver prompt action on harmful content or activity.”

“Rather than requiring the removal of specific pieces of legal content, regulation will focus on the wider systems and processes that platforms have in place to deal with online harms, while maintaining a proportionate and risk-based approach.”

“In fact, the new regulatory framework will not require the removal of specific pieces of legal content. Instead, it will focus on the wider systems and processes that platforms have in place to deal with online harms, while maintaining a proportionate and risk-based approach.”

“Of course, companies will be required to take particularly robust action to tackle terrorist content and online Child Sexual Exploitation and Abuse. The new regulatory framework will not remove companies’ existing duty to remove illegal content.”

The Ministers said:

Caroline Dinenage: “the codes of practice are really about systems and processes, rather than naming individual harms in the legislation. There are two exceptions to that: there will be codes of practice around child sexual exploitation and terrorist content, because those are both illegal.” (Q554)

“It is for the regulator to set out codes of practice, but they won’t be around individual harms; they will be around systems and processes—what we expect the companies to do. Rather than focusing on individual harms, because we know that the technology moves on so quickly that there could be more, it is a case of setting out the systems and processes that we would expect companies to abide by, and then giving the regulator the opportunity to impose sanctions on those that are not doing so.” (Q.556)

Q562 Stuart C. McDonald: “…if the regulator feels that algorithms are working inappropriately and directing people who have made innocent searches to, say, far-right content, will they be able to order, essentially, the company to make changes to how its algorithms are operating?

Caroline Dinenage: Yes, I think that they will. That is clearly something that we will set out in the full response. The key here is that companies must have clear transparency, they must set out clear standards, and they must have a clear duty of care. If they are designing algorithms that in any way put people at risk, that is, as I say, a clear design choice, and that choice carries with it a great deal of responsibility. It will be for the regulator to oversee that responsibility. If they have any concerns about the way that that is being upheld, there are sanctions that they can impose.”

The Consequence: As with the specific issue around the status of terms and conditions for “lawful but harmful” content (see below), it is difficult to see how a bright line can be drawn between substance and process.  Processes cannot be designed, risk-assessed or their effectiveness evaluated in the abstract — only by reference to goals such as improving user safety and reducing risk of harm. A duty of care evaluated without reference to the kind of harm intended to be guarded against makes no more sense than the smile without the Cheshire Cat. 

In Caparo v Dickman Lord Bridge cautioned against discussing duties of care in the abstract:
"It is never sufficient to ask simply whether A owes B a duty of care. It always necessary to determine the scope of the duty by reference to the kind of damage from which A must take care to save B harmless."

Risk assessment is familiar in the realm of safety properly so-called: danger of physical injury, where there is a clear understanding of what constitutes objectively ascertainable harm. It breaks down when applied to undefined, inherently subjective harms arising from users' speech. If "threatening public confidence in the 5G roll-out” (see above) can be labelled an online harm within scope of the legislation, that goes far beyond any tenable concept of safety.

The government’s approach appears to be to adopt different approaches to illegal and “legal but harmful”, the latter avowedly restricted to process (although see next topic as to how far that can really be the case). 

In passing, the Initial Response is technically incorrect in referring to “companies’ existing duty to remove illegal content”. No such general duty exists. Hosting providers lose the protection of the ECommerce Directive liability shield if they do not remove unlawful content expeditiously upon gaining actual or (for damages) constructive knowledge of the illegality. Even then, the eCommerce Directive does not oblige them to remove it. The consequence is that they become exposed to the risk of possible liability (which may or may not exist) under the relevant underlying law (see here for a fuller explanation). In practice that regime strongly incentivises hosting providers to remove illegal content upon gaining relevant knowledge. But they have no general legal obligation to do so.

3. For “lawful but harmful” content seen by adults, will the regulator be interested only in whether intermediaries are enforcing whatever content standards they choose to put in their TandCs?

The White Paper said:

“As indication of their compliance with their overarching duty of care to keep users safe, we envisage that, where relevant, companies in scope will:

  • Ensure their relevant terms and conditions meet standards set by the regulator and reflect the codes of practice as appropriate.
  • Enforce their own relevant terms and conditions effectively and consistently. …”
“To help achieve these outcomes, we expect the regulator to develop codes of practice that set out: 

  • Steps to ensure products and services are safe by design.
  • Guidance about how to ensure terms of use are adequate and are understood by users when they sign up to use the service. …
  • Steps to ensure harmful content or activity is dealt with rapidly. …
  • Steps to monitor, evaluate and improve the effectiveness of their processes.”
The Initial Response said:

“We will not prevent adults from accessing or posting legal content, nor require companies to remove specific pieces of legal content. The new regulatory framework will instead require companies, where relevant, to explicitly state what content and behaviour is acceptable on their sites and then for platforms to enforce this consistently.”

“To ensure protections for freedom of expression, regulation will establish differentiated expectations on companies for illegal content and activity, versus conduct that is not illegal but has the potential to cause harm. Regulation will therefore not force companies to remove specific pieces of legal content. The new regulatory framework will instead require companies, where relevant, to explicitly state what content and behaviour they deem to be acceptable on their sites and enforce this consistently and transparently. All companies in scope will need to ensure a higher level of protection for children, and take reasonable steps to protect them from inappropriate or harmful content.”

“Recognising concerns about freedom of expression, the regulator will not investigate or adjudicate on individual complaints. Companies will be able to decide what type of legal content or behaviour is acceptable on their services, but must take reasonable steps to protect children from harm. They will need to set this out in clear and accessible terms and conditions and enforce these effectively, consistently and transparently.”

The Ministers said:

Oliver Dowden: “The essence of online harms legislation is holding social media companies to what they have promised to do and to their own terms and conditions. My focus in respect of those is principally on two things: underage harms and illegal harms. Clearly, the trickiest category is legal adult harms. In respect of that, we are looking at how we tighten the measures to ensure that those companies actually do what they promised they would do in the first place, which often is not the case.” (Q20) (emphasis added)

“Clearly, in respect of legal adult harms, that is the underlying principle anyway in the sense that what we are really trying to do is say to those social media companies and tech firms, “Be true to what you say you are doing. Just stick by your terms and conditions”. We would ask the regulator to make sure that it is enforcing them, and then have tools at our disposal to require it to do so.” (Q89) (emphasis added)

Caroline Dinenage: “A lot of this is about companies having the right regulations and standards and duty of care, and that will also be in the online harms Bill and online harms work. If we can have more transparency as to what platforms regard as acceptable—there will be a regulator that will help guide them in that process—I think we will have a much better opportunity to tackle those things head-on.” (Q513) (emphasis added)

“With regard to our role in DCMS, it is more as a co-ordinator bringing together the work of all the different Government Departments and then liaising directly with the platforms to make sure that their standards, their regulations, are reflective of some of the concerns that we have—make sure, in some cases, that harmful content can be anticipated and therefore prevented, and, where that is not possible, where it can be stopped and removed as quickly as possible.” (emphasis added) (Q525)

Baroness Williams: “There is obviously that which is illegal and that which breaches the CSPs’ terms of use. It is that latter element, particularly in the area of extremism, on which we have really tried to engage with CSPs to get them to be more proactive.” (emphasis added) (emphasis added) (Q.527)

The Consequence: This is now one of the most puzzling areas of the government’s developing policy. The White Paper expected that codes of practice would ensure that terms and conditions meet “standards set by the regulator” and that terms of use are “adequate”. These statements were not on the face of them limited to procedural standards and adequacy. They could readily be interpreted as encompassing standards and adequacy judged by reference to harm reduction goals determined by the regulator (which, as we have seen, would be able to decide for itself what constitutes harm) – in other words, extending to the substantive content of intermediaries' terms and conditions.

When the Initial Response was published, great play was made of the shift to a differentiated duty of care: that it would be up to the intermediary to decide – for lawful content for adults - what standards to put in its terms and conditions. 

The remit of the regulator would be limited to ensuring those standards are clearly stated and enforced “consistently and transparently” (or “effectively, consistently and transparently”, depending on which part of the Initial Response you turn to; or “effectively and consistently”, according to the White Paper). Indeed the Secretary of State said in evidence that "The essence of online harms legislation is holding social media companies to what they have promised to do and to their own terms and conditions

But it seems from the other Ministers’ responses that the government has not disclaimed all interest in the substantive content of intermediaries’ terms and conditions. On the contrary, the government evidently sees it as part of its role to influence (to put it at its lowest) what goes into them. If the regulator’s task is to ensure enforcement of terms and conditions whose substantive content reflects the wishes of a government department, that is a far cry from the proclaimed freedom of intermediaries to set their own standards of acceptable lawful content.

Ultimately, what can be the point of emphasising how, in the name of upholding freedom of speech, the role of an independent regulator will be limited to enforcing the intermediaries’ own terms and conditions, if the government considers that part of its own role is to influence those intermediaries as to what substantive provisions those TandCs should contain?

This is one aspect of an emerging issue about division of responsibility between government and the regulator. It is tempting to think that once an independent regulator is established the government itself will withdraw from the fray. But if that is not so, then reducing the remit of the independent regulator concomitantly increases the scope for the government itself to step in.

That is especially pertinent in the light of the government’s desire to cast itself as a ‘trusted flagger’, whose notifications of unlawful content the intermediaries should act upon without question. Thus Caroline Dinenage appears to regard the platforms as obliged to remove anything that the government has told them it considers to be illegal (with no apparent requirement of prior due process such as independent verification), and would like them to take seriously anything else that the government notifies to them:

“We have found that we have become—I forget the proper term, but we have become like a trusted flagger with a number of the online hosting companies, with the platforms. So when we flag information, they do not have to double-check the concerns we have. Clearly, unless something is illegal, we cannot tell organisations to take it down; they have to make their own decision based on their own consciences, standards and requirements. But clearly we are building up a very strong, trusted relationship with them to ensure that when we flag things, they take it seriously.” (Emphasis added)

4. Codes of Practice for specific kinds of user content or activity?

The White Paper said:

“[T]he White Paper sets out high-level expectations of companies, including some specific expectations in relation to certain harms. We expect the regulator to reflect these in future codes of practice.”

It then set out a list of 11 harms, accompanied in each case by a list of areas in relation to that harm that it expected the regulator to include in a code of practice. For instance, in relation to disinformation a list of 11 specific areas included:

“Steps that companies should take in relation to users who deliberately misrepresent their identity to spread and strengthen disinformation.”; and

“Promoting diverse news content, countering the ‘echo chamber’ in which people are only exposed to information which reinforces their existing views.”

The Initial Response said:

“The White Paper talked about the different codes of practice that the regulator will issue to outline the processes that companies need to adopt to help demonstrate that they have fulfilled their duty of care to their users. … We do not expect there to be a code of practice for each category of harmful content, however, we intend to publish interim codes of practice on how to tackle online terrorist and Child Sexual Exploitation and Abuse (CSEA) content and activity in the coming months.”

The Ministers said:

Caroline Dinenage: I think I need to clear up a bit of a misunderstanding about the White Paper. The 11 harms that were listed were really intended to be an illustrative list of what we saw as the harms. The response did not expect a code of practice for each one, because the codes of practice are really about systems and processes, rather than naming individual harms in the legislation. There are two exceptions to that: there will be codes of practice around child sexual exploitation and terrorist content, because those are both illegal.” (Q.554) (emphasis added)

The Consequence: The different approach to CSEA and terrorism probably owes more to the different areas of responsibility of the Home Office and the DCMS than to any dividing line between illegality and non-illegality. The White Paper covers many more areas of illegality than those two alone.

5. Search engines in scope?

The White Paper said:

“… will apply to companies that allow users to share or discover user-generated content, or interact with each other online.” (emphasis added)

“These services are offered by…  search engines” (Executive Summary)

The Initial Response said:

“The legislation will only apply to companies that provide services or use functionality on their websites which facilitate the sharing of user generated content or user interactions, for example though comments, forums or video sharing” (emphasis added)

The Ministers said:

Caroline Dinenage: Again, we are probably victims of the fact that we published an interim response, which was not as comprehensive as our full response will be later on in the year. The White Paper made it very clear that search engines would be included in the scope of the framework and the nature of the requirements will reflect the type of service that they offer. We did not explicitly mention it in the interim response, but that does not mean that anything has changed. It did not cover the full policy. Search engines will be included and there is no change to our thoughts and our policy on that.”   (Q.560)

The Consequence: Notwithstanding the Minister’s explanation, the alterations in wording between the White Paper and the Initial Response (omitting “discover”, adding “only”) had the appearance of a considered change. The lesson for the future is perhaps that it would be unwise to parse too closely the text of anything else said or written by the government.

6. Everything from social media platforms to retail customer review sections?

The White Paper said:

“… companies of all sizes will be in scope of the regulatory framework. The scope will include… social media companies, public discussion forums, retailers that allow users to review products online, along with non-profit organisations, file sharing sites and cloud hosting providers.” (emphasis added)

The Initial Response said:

“To be in scope, a business would have to operate its own website with the functionality to enable sharing of user-generated content, or user interactions.”

The Ministers said:

Oliver Dowden: “We are a Europe leader in this. I have seen, as I am sure you have seen, the unintended consequences of good-intended legislation then having bureaucratic implications and costs on businesses that we want to avoid.

For example, in respect of legal online harms for adults, if you are an SME retailer and you have a review site on your website for your product and people can put comments underneath that, that is a form of social media. Notionally, that would be covered by the online harms regime as it stands. The response to that is they will go through this quick test and then they will find it does not apply to them. My whole experience of that for SMEs and others is that it is all very well saying that when you are sat have no idea what this online harms thing is, this potentially puts a big administrative burden on you. (emphasis added)

Are there ways in which we can carve out those sorts of areas so we focus on where we need to do it? Those kinds of arguments pertain less to illegal harms and harms to children. I hope that gives you a flavour of it.” (Q.88)

Q89 Damian Hinds: “Yes, quite so. I think in the previous announcement there was quite a high estimate of the number of firms or proportion of total firms that would somehow be counted in the definition of an online platform, which was rather a disturbing thought. It would be very welcome, what you can do to limit the scope of who counts as a social media platform.”

The Consequence: This exchange does shine a light on the expansive scope of the proposed legislation. The Secretary of State said that SME retailers with review sections were “notionally” covered. However, there was nothing notional about it.  Retailer review sections were expressly included in the White Paper, as were companies of all sizes.

As the Secretary of State suggests, it is little comfort for an SME to be told “don’t worry, you’ll be low risk so it won’t really apply to you” if: (a) you are in scope on the face of it, and (b) it is left to the regulator to decide whether the duty of care should bear less heavily on some intermediaries than others. 

There are, of course, many other kinds of non-social media platform intermediary who are in scope as well as SME retailers with review sections: apps, online games, community discussion forums, non-profits and many other online services.  The Initial Response said “Analysis so far suggests that fewer than 5% of UK businesses will be in scope of this regulatory framework.” Whether 5% is considered to be small or large in absolute terms (not to mention the apparent indifference to non-UK businesses), there has been no indication of the assumptions underlying that estimate.

7. Will journalism and the press be excluded from scope?

The White Paper said:

Nothing. In a subsequent letter to the Society of Editors the then DCMS Secretary of State Jeremy Wright said:

“… as I made clear at the White Paper launch and in the House of Commons, where these services are already well regulated, as IPSO and IMPRESS do regarding their members' moderated comment sections, we will not duplicate those efforts. Journalistic or editorial content will not be affected by the regulatory framework.”

The Initial Response said:

Nothing. It limited itself to general expressions of support for freedom of expression, such as:
“…freedom of expression, and the role of a free press, is vital to a healthy democracy. We will ensure that there are safeguards in the legislation, so companies and the new regulator have a clear responsibility to protect users’ rights online, including freedom of expression and the need to maintain a vibrant and diverse public square.”

The Ministers said:

Caroline Dinenage: Obviously, we know that a free press is one of the pillars of our society, and the White Paper, I must say from the outset, is not seeking to prohibit press freedom at all, so journalistic and editorial content is not in the scope of the White Paper. Our stance on press regulation has not changed.” (Emphasis added)

“As for what has been in the papers recently, the Secretary of State wrote a letter to the Society of Editors, and this was about what you might call the below-the-line or comments section. They were concerned that that might be regulated. I think what the Secretary of State is saying is that, where there is already clear and effective moderation of that sort of content, we do not intend to duplicate it. For example, there is IPSO and IMPRESS activity on moderated content sections. Those are the technical words for it. This is still an ongoing conversation, so we are working at the moment with stakeholders to develop proposals on how we are going to reflect that in legislation, working around those parameters. (Q.558)

“Stuart C. McDonald: But there is no suggestion that below-the-line remains unregulated. It is where that regulation should lie that is the issue.

Caroline Dinenage: Exactly.” (Q.559)

The Consequence: There are three distinct issues around inclusion or exclusion of the press from the regulatory scope of the Bill:

1. User comments on newspaper websites.  On the face of it, news organisations would be subject to the duty of care as regards user comments on their websites. The position of the government appears to be that whether the duty of care would apply would depend on whether the comments are already subject to another kind of regulation (or at least the existence of “clear and effective moderation”). Potentially, therefore, newspapers that are not regulated by IPSO or IMPRESS would be in scope for this purpose. Whether this demarcation would be achieved by a hard scope exclusion written into the Bill is not clear.

2. Journalistic or editorial material. Whilst the Minister may say that the government’s stance on press regulation has not changed, her statement that journalistic and editorial content is not “in the scope” of the White Paper is new — at least if we are to understand that as meaning that the Bill would contain a hard scope exclusion for journalistic or editorial content. Previously the government had said only that such content would not be affected by the regulatory framework. A general exclusion of journalistic or editorial material would on the face of it go much wider than newspapers and similar publications. It would be no surprise to find this statement being “clarified” at some point in the future.

3. Newspaper social media feeds and pages. Newspapers and other publications maintain their own pages, feeds and blogs on social media and other platforms. Newspapers would not themselves be subject to a duty of care in relation to their own content. But as far as the platforms are concerned the newspapers are users, so that their pages and feeds would fall under the platforms’ duty of care. As such, they would be liable to have action taken against their content by a platform in the course of fulfilling its own duty of care.

The government has said nothing about whether, and if so how, such press content would be excluded from scope. If the government is serious about excluding “journalistic or editorial” material generally from scope, that would achieve this. However that would create immense difficulties around whether a particular feed or page is or is not journalistic or editorial material (what about this Cyberleagle blog, or the Guido Fawkes blog, for instance?), and how a platform is supposed to decide whether any particular content is or is not in scope.  

8. End to end encryption

The White Paper said:

Nothing. (Although the potential for the duty of care to be applied to prevent the use of end to end encryption was evident.)

The Initial Response said:


The Ministers said:

Baroness Williams: “[Facebook] then announced that they were going to end-to-end encrypt Messenger. That, for us, is gravely worrying, because nobody will be able to see into Messenger. I know there is going to be a Five Eyes engagement next week, and I do not know if the Committee knows, but the Five Eyes wrote to Mark Zuckerberg last year, so worried were we about this development.” (Q538)

Q566 Chair: “On that basis, does end-to-end encryption count as a breach of duty of care?

Baroness Williams:It is criminal activity that would breach the duty of care. Allowing criminal activity to happen on your platform would be the breach of duty of care. End-to-end encryption, in and of itself, is not a breach of duty of care.

Chair: Presumably, for this regulation to have any bite at all, they will have to be able to take some enforcement against the policies that fail to prevent criminal activity. On that logic, introducing the end-to-end encryption, if it knowingly stops the company from preventing illegal activity—for example, the kind of online child abuse you have talked about—that would surely count as a breach of duty of care.

Baroness Williams: I fully expect that that is what some of the Five Eyes discussions, which will be happening very shortly, will look at.”

The Consequence: This is the first indication that the government is alive to the possibility that a regulator might be able to interpret a duty of care so as to affect the ability of an intermediary to use end to end encryption. The “in and of itself” phraseology used by the Minister appears not to rule that out. This issue is related to the question of how the legislation might apply to private messaging providers, a topic on which the government has consulted but has not yet published a conclusion.

9. Identity verification

The White Paper said:

“The internet can be used to harass, bully or intimidate. In many cases of harassment and other forms of abusive communications online, the offender will be unknown to the victim. In some instances, they will have taken technical steps to conceal their identity. Government and law enforcement are taking action to tackle this threat.”

“The police have a range of legal powers to identify individuals who attempt to use anonymity to escape sanctions for online abuse, where the activity is illegal. The government will work with law enforcement to review whether the current powers are sufficient to tackle anonymous abuse online.”

“Some of the areas we expect the regulator to include in a code of practice are:

  • Steps to limit anonymised users abusing their services, including harassing others. …
  • Steps companies should take to limit anonymised users using their services to abuse others.”

The Initial Response said:


The Ministers said:

Q25 John Nicolson: Would you like to see online harms legislation compel social media companies to verify the identity of users, not of course to publish them but simply to verify them before the accounts are up and running?

Oliver Dowden: There is certainly a challenge around, as you mentioned, bots, which are sometimes used by hostile state activity, and finding better ways of verifying to see whether these are genuine actors or whether it is co-ordinated bot-type activity. That is through online harms but there is obviously a national security angle to that as well.”

Q530 Ms Abbott: “Finally, would you consider changing the regulation, so you could post anonymously on a website or Twitter or Facebook, but the online platform would have your name and address? In my experience, when you try to pursue online abuse, you hit a brick wall because the abuser is not just anonymous when they post, the online platform doesn’t have a name and address either.

Caroline Dinenage: That is a really interesting idea. It is definitely something that we have been discussing. With regard to the online harms legislation that we are putting together at the moment, we have said very clearly that companies need to be much more transparent. They need to set out standards and they need to clarify what their duty of care is and to have a robust complaints procedure that people can use and can trust in. That is why we are also appointing a regulator that will set out what good looks like and will have expectations but also powers to be able to demand data and information and to be able to impose sanctions on those that they do not feel are abiding by them.

Q531 Chair: What does that actually mean? Does that mean that you think that the regulator should have the power to say that social media companies should not allow people to be … [a]nonymous to the platform?

Caroline Dinenage: This is something that we are considering at the moment. There are a number of things here. In the online harms legislation, the regulator will set out their expectations.

Chair: We can’t devolve everything to the regulator. Something like this is really important—should social media companies be allowed to not know who it is that is using their platforms? That feels like a big question that Parliament should take a view on, not something we just hand over to a regulator and say, “Okay, whatever you think,” later on.

Caroline Dinenage: Yes, exactly. That is why we are considering it at the moment, as part of the online harms legislation, and that, of course, will come before Parliament.”

Q545 Tim Loughton: “… If I want to set up a bank account and all sorts of other accounts, I must prove to the bank or organisation who I am by use of a utility bill and other things like that. It is quite straightforward. What is the downside of a similar requirement being enforced by social media platforms before you are allowed to sign up for an account? This is an issue that we have looked at before on the Committee. Many of us have suggested that we should go down that route. I gather that it already happens in South Korea. You say that you are looking at it, Minister Dinenage. What, in your view, is the downside of having such scrutiny?

Caroline Dinenage: You make a very compelling argument, Mr Loughton. A lot of what you said is extremely correct. The only thing we are mulling over and trying to cope with is whether there is any reason for anonymity for people who are victims, who want to be able to whistleblow, and who may be overseas and might not want to identify themselves because they fear for their lives or other harm. There are those issues of anonymity and protecting someone’s safety and ability to speak up. That is what we are wrestling with.

Q546 Tim Loughton: By the same token, you could have somebody with a fake identity who is falsely whistleblowing or pushing around propaganda, so it cuts both ways. I fail to see the downside of having a requirement that you have to prove who you are—not least because we know what happens when people are caught and have their sites taken down. Five minutes later, they set up another new anonymous site peddling the same sort of false information.

Caroline Dinenage: You make a very compelling argument. This is such an important piece of legislation, and we have to get it right. As I say, it is world-leading. Everybody is looking at us to see how we do it. We need to make sure that we have taken into consideration every angle, and that is what we are doing at the moment.”

The Consequence: Identity verification is evidently an issue that is bubbling to the surface. The most fundamental objection is that the right of freedom of expression secured by Article 19 of the Universal Declaration of Human Rights is not conditioned upon identity verification. It does not say:

"Everyone has the right to freedom of opinion and expression upon production of any two of the following: driving licence, passport, recent utility or council tax bill...".

In South Korea, legislation imposing online identity verification obligations was declared unconstitutional in 2012.

The Home Affairs Committee raised, to the best of my knowledge for the first time in any Parliamentary deliberation on the Online Harms project, the question of what should be decided by Parliament and what delegated to a regulator. That is not limited to the question of identity verification. It is an inherent vice of regulatory powers painted with such a broad brush that many concrete issues will lie hidden behind abstractions, to surface only when the regulator turns its light upon them – by which time it is far too late to object that the matter should have been one for Parliament to decide. That vice is compounded when the powers affect the individual speech of millions of people.

10. Extraterritoriality

The White Paper said:

“The new regulatory regime will need to handle the global nature of both the digital economy and many of the companies in scope. The law will apply to companies that provide services to UK users.” (6.9) (emphasis added)

“We are also considering options for the regulator, in certain circumstances, to require companies which are based outside the UK to appoint a UK or EEA-based nominated representative.” (6.10)

The Initial Response said:

Nothing of relevance.

The Ministers said:

“Q569: Andrew Gwynne: Presumably the regulations will apply to all content visibly available in the UK—is that correct?

Baroness Williams: Yes.”

The Consequence: Charitably, perhaps we should assume that the Minister misspoke. There is a vast difference between providing services to UK users and mere visibility in the UK. Given the inherent cross-border nature of the internet, asserting a country’s local law against content on a mere visibility basis is tantamount to asserting world-wide extra-territoriality. 

It would be more consistent with the direction in which internet jurisdictional norms have moved over the last 25 years to apply a test of whether the provider is targeting the UK.

Sunday 10 May 2020

Decrypting eIDAS

The EU’s eIDAS (Electronic Identification, Authentication and Trust Services) Regulation was launched in October 2014, ushering in – it was hoped - a new era of digital signing, validated documents, secure electronic document delivery, verified time-stamps, authenticated websites and intra-EU cross-border recognition of signatures.

In large part (it also covers Member State public sector electronic identity systems) eIDAS is version 2 of an EU initiative that started 20 years ago, with the 1999 Electronic Signatures Directive.

At the eIDAS launch ceremony the outgoing EU Digital Commissioner Neelie Kroes signed a letter to President-Elect Jean-Claude Juncker using a “qualified signature”, the most secure variety of digital signature defined by eIDAS. She said.

“I am confident that I have laid the foundation for you to build a digitally-strong house. The eIDAS Regulation was the missing stone to make cross-border electronic transactions across Europe a reality.”

Some geo-political tub-thumping was also evident:

“With eIDAS we have accomplished a major milestone - and we are well ahead of the US in this.”

Finally, the signature:

As I like to practise what I preach, I am signing this letter electronically, with my mobile, and using technology developed thanks to the EU funded STORK project which is currently used by citizens in Austria.”

The eIDAS ecosystem

The launch marked the official start of a project to create an eIDAS “ecosystem”. The project aims at fostering wider adoption of digital signatures: defined at a high level by the Regulation, then given substance by technical standards promulgated under its umbrella. The Regulation, adopted in July 2014, came fully into force in July 2016.

The eIDAS ecosystem is populated by a menagerie of signature generation services, certification authorities, time stamping services, validation services, preservation services, users and others. Nearly 40 pages of the Regulation (plus the associated volumes of technical standards and guidance) aim to tear down the barriers that — so it is said — stand in the way of signing, sealing and delivering documents electronically.

eIDAS – central or peripheral?

One might expect, then, that eIDAS would be front and centre when we analyse the ability to use electronic signatures under English law. All the more so since, unlike its predecessor Directive, as an EU Regulation eIDAS is directly incorporated into English law. But not so. Counter-intuitively, eIDAS sits on the sidelines and performs little more than a supporting role.

Direct incorporation as an EU Regulation does revive some interpretative riddles that, during the period of the Directive, could be left unresolved. Whatever the Directive might have meant, there was no doubt that English law complied with it. Those riddles are harder to ignore now that they form part of English law. Later on, this piece attempts to crack them.

The conclusion, thankfully, is that the eIDAS riddles can and should be solved in a way that leaves the liberal and facilitative English law of signatures untouched.

That conclusion sets the scene for the first major topic of this piece: how can it be that, when we analyse the ability to use electronic signatures from an English law perspective, the elaborately constructed edifice of eIDAS turns out to be more decorative grotto than grand mansion?
The high level answer is that the eIDAS framework, which defines some specific technical categories of signature, is swallowed up in the broader English regime under which an electronic signature of any kind (including something as informal as typing a name or initials at the end of an email) can count as a signature. Consequently the technical categories of signature defined by eIDAS (and the Directive before it) have little legal significance.

If the law does not require the use of an eIDAS-compliant “advanced signature” (very rarely in English statutes) or “qualified signature” (no instances in English statutes), and if the law does not confer any exclusive status on such eIDAS-compliant signatures (as to which more below), then people are free to choose the kind of electronic signature that suits the transaction in which they are about to engage.

That, since the Directive came into force 20 years ago, is what they have done. If users adjudge that they do not require the high levels of assurance as to identity and data integrity aimed at by EU standards, they are unlikely to pay a premium for expensively engineered and supported standards-compliant cryptographic signature products. 

How did the Directive, and subsequently eIDAS, come to adopt the approach that they have?

Reverence for the handwritten signature

One clue lies in reverence for the assumed properties of a handwritten signature. Underlying both eIDAS and its predecessor Directive is the implicit assumption that the handwritten signature offers a high degree of protection against forgery and provides a strong physical connection to the signed document, both of which characteristics should be replicated in the electronic world.

Such reverence is understandable to some extent if one thinks only of a full name, distinctively styled, inscribed indelibly in ink on paper: the raw material on which a forensic handwriting expert can work if necessary. But that would be to misdescribe English law (although perhaps not that of some other countries that have traditionally set great store by the observance of formalities). The English law of signatures has not required perfection or anything approaching it. Otherwise English law would not have permitted, as it has done, an ‘X’ or a facsimile rubber stamp to count as a signature.  That liberal approach to physical signatures sets the tone for the English law approach to electronic signatures.

Reverence for the handwritten signature is a likely source of the assumption underlying the Directive and eIDAS that the electronic functional equivalent of a handwritten signature is a cryptographic digital signature tied to a third party certificate, providing a high level of confidence that the signatory is who they purport to be and that the signed document has not been tampered with. Lack of such confidence is said to undermine trust in the digital environment, creating an impediment to electronic transactions.

Such comparisons, however, rely on deeper (not necessarily well-founded) assumptions about the practical function and significance of a physical signature, the degree of assurance that users expect or require from any kind of signature, and the status of a signature in national law.  

As to problems of interpretation, eIDAS relies on some definitional concepts that, whilst simple on their face, are nevertheless elusive. For the eiDAS community itself this may not always be an issue. If products and services comply with the Commission’s promulgated technical standards they are presumed to comply with the Regulation’s definitions, whatever those may mean. But for anyone trying to analyse and apply eIDAS in a wider context, the Regulation and its predecessor Directive present considerable interpretative challenges.

What does a signature do?

Before grappling with the interpretative riddles of eIDAS, let us consider the function and legal significance of a signature.

A signature may from a legal perspective be considered to have three[1] functions, each of which may be present to a greater or lesser degree:

  • Identification of the signatory
  • Demonstrating the signatory’s intention to be bound by, or at least adopt, the contents of the document
  • Identification of the contents of the signed document
All three functions can be thought of as aspects of non-repudiation: preventing the signatory from denying that they signed the document at all, from denying that they intended to be bound by it, or from denying that they signed a document in those terms.

However, we run into trouble if we turn this round and make effectiveness at achieving non-repudiation the sine qua non of legal recognition as a signature. Doing so risks losing sight of the extent to which traditional signatures fall short of guaranteeing non-repudiation, but nevertheless are recognised in law as functional signatures.

The observation of the Australian Electronic Commerce Expert Group in its March 1998 Report to the Attorney-General is apposite:

‘‘There is always the temptation, in dealing with the law as it relates to unfamiliar and new technologies, to set the standards required of a new technology higher than those which currently apply to paper and to overlook the weaknesses that we know to inhere in the familiar.’’

As the Law Commission noted in its 2001 Advice to Government:
“English law has long accepted a ‘signature’ in the form of an ‘X’ though this does not identify the ‘signatory’ in any real sense.”

Ultimately it is the second function – demonstrating an intention to be bound – that for English law purposes is the defining characteristic of a signature. Put more broadly, that function can be stated as an intention to adopt the contents of the document (or perhaps part of it) or to attribute legal significance to it.

This function is often described as demonstrating an authenticating intention on the part of the signatory. The Law Commission, in its September 2019 Report on Electronic Execution of Documents, concluded that:

“An electronic signature is capable in law of being used to execute a document1 (including a deed) provided that (i) the person signing the document intends to authenticate the document and (ii) any formalities relating to execution of that document are satisfied.”

The term "authentication" is, however, liable to confuse.

It can be understood to mean assuring the identity of the signatory or the contents of the signed document — in other words the first and third possible functions of a signature  respectively.  eIDAS now defines ‘authentication’ in that way: as an electronic process that “enables the electronic identification of a natural or legal person, or the origin and integrity of data in electronic form to be confirmed”.

The 2019 Law Commission Report points out that, by contrast, English law uses authentication to mean the second function:

“3.29 … We use this language because this is what is used in case law. What it means, effectively, is that the party intended to be bound by the document.”

The Directive’s definition of an electronic signature required that it “serve as a method of authentication”. However, since authentication was left undefined it was unclear in what sense it was being used. eIDAS, as seen above, has now introduced a definition of authentication in terms of identification of signatory and data. However, it no longer stipulates authentication as a defining characteristic of an electronic signature. That definition has moved towards an English law approach: “data in electronic form which is … used by the signatory to sign.”

Technologists and policy-makers may tend to assume that an electronic signature cannot be valid (or at least useful) unless it performs each of the three non-repudiation functions to a high level of confidence. Even when specified in a technology-neutral fashion, that tends to lead to complex schemes involving third party certification of the signatory’s identity, allied with cryptographic methods of securing the signature and of demonstrating that the document has not been altered.

Such techniques, however, go far beyond achieving equivalence with the capabilities of an ordinary handwritten signature. To the extent that a traditional signature performs the first and third functions (identification of signatory and document), it may do so only weakly.

Even a full handwritten autograph is not proof against forgery. An ‘X’ marked at the end of the document (which in English law is capable of operating as a signature) barely, if at all, identifies the signatory. Neither it nor an autograph signature infallibly identifies the contents of the document (at least, one consisting of several pages), nor renders it tamper-evident. To the extent that digital signatures have sought to render electronically signed documents tamper-evident, that emulates the qualities of paper rather than that of the signature inscribed upon it.

A historic quest for near-perfect non-repudiation mechanisms lies behind some policy and technological developments in the electronic signatures field. A tendency to prescribe technology-specific requirements was especially evident in some 1990s legislation (the 1995 Utah Digital Signature Act being the paradigm example). Whilst legislatures have generally moved on since then towards adopting more technology-neutral approaches, the echoes still reverberate.

Now let us turn from function to legal significance of a signature.  On most occasions when we sign documents the law does not require a signature to be used. Sometimes, however, it does so. Compliance with such a mandatory requirement is a different and separate issue from the legal significance of a signature generally, which is discussed below. 

Mandatory signatures

If the law (usually a statute) does require a signature to be used in a particular situation, then does the kind of signature proposed to be used comply with that requirement? The answer is likely to affect the validity of the document or of a transaction to which it relates.

eIDAS does not prevent a Member State from enacting legislation that stipulates formalities that have to be complied with for a particular purpose, including use of a particular kind of signature (whether wet-ink or electronic) (see discussion under ‘Legal Effect’ below).

The 2019 Law Commission Report observed that there is an argument that eIDAS would have allowed the common law to develop to the effect that an electronic signature was not a valid way of signing a contract.

However, English common law has not done that. Conclusion of a contract is not generally subject to a statutory requirement for a signature.

Furthermore, English law has taken a liberal view of what constitutes a signature for the purposes of a generally expressed statutory signature requirement, encompassing all kinds of electronic signature including the most informal. In English law any kind of electronic signature is capable of satisfying a generally expressed requirement for a signature, so long as there is an intention thereby to adopt the contents of the document and so long as any other applicable formalities are satisfied (see 2019 Law Commission Report, above).

Non-mandatory signatures

Signing a document without any statutory or other legal requirement for a signature may (or may not) have some legal significance. Thus:

  • signature is one way of concluding a contract and indicating assent to its terms (even though the law does not in general require a contract to be signed).
  • when we sign a letter the signature associates us with the final contents. That may have legal consequences (if, for instance, the letter provides a reference on which the recipient will rely).
  • if we sign a painting we do so in order to adopt it as our work.
  • if we sign a receipt we are acknowledging that the goods, services or money have been received.
  • if we sign a document as a witness, we do so to indicate that we observed the signatory signing the document (but we do not thereby endorse the contents of the document).
  • if a celebrity signs an autograph book, they are providing a specimen of their signature. There is no intent, by signing the book, to adopt any of its contents other than the signature itself.
Thus the legal effect (if any) of a signature may vary considerably depending on the purpose for which the signature is applied and the context in which it is later relied upon.
The riddles of eIDAS

Now let us turn to eIDAS. eIDAS describes several kinds of electronic signature. Most relevantly, it defines advanced electronic signatures and qualified electronic signatures.

What those consist of in technical terms need not detain us for the moment. Suffice to say that both attempt to incorporate all three signature functions (identity of signatory, intention to authenticate and identification of document contents) to a high level of confidence. Advanced signatures can be thought of as the eIDAS silver standard, qualified signatures as the gold standard. The bronze standard is any other signature in electronic form.

eIDAS itself does not compel the use of advanced or qualified signatures. Nor does it require Member States to compel private parties to use either of them in their dealings. They are in the nature of prefabricated tools that Member States (or indeed private parties in contracts) are at liberty to prescribe if they wish to do so.

So where lies the problem of interpretation? The EU legislature wanted to do two further things: (a) prevent Member States from discriminating against use of electronic signatures as such and (b) attribute a specific status to its gold standard “qualified signature”. But at the same time Member States were to be free to lay down whatever formalities for transactions between private parties they saw fit. That could include stipulating that a specific kind of signature had to be used in any given situation.

The attempt to reconcile these objectives resulted first in in Article 5 of the Directive, then Article 25 of eIDAS. Article 25 states (in part):
“1. An electronic signature shall not be denied legal effect and admissibility as evidence in legal proceedings solely on the grounds that it is in an electronic form or that it does not meet the requirements for qualified electronic signatures.”
2.   A qualified electronic signature shall have the equivalent legal effect of a handwritten signature. …”

Article 5 of the Directive employed broadly comparable terminology: “legal effectiveness” and “satisfy the legal requirements of a signature in electronic form in the same manner as a handwritten signature satisfies those requirements in relation to paper-based data”.

Legal effect and admissibility in evidence are, on the face of them, separate concepts. Admissibility in evidence is simple: the court is able to look at the signature (although it is still free to give it as much or as little evidential weight as it thinks fit).  Legal effect, however, is a puzzle.  

Riddle 1: Legal Effect

‘Legal effect’ is mentioned in both Article 25.1 and 25.2 of eIDAS.

Article 25.1 is the non-discrimination provision.  Legal effect must not be denied either solely on grounds of electronic form or because it is not a gold-standard qualified signature.

But what constitutes legal effect? At first sight it could mean that any electronic signature must be taken to satisfy a national law statutory requirement for a signature.  However, that cannot be the answer, for several reasons.

Purposively, it would be a pointless exercise for the Regulation to define different kinds of electronic signature if Article 25(1) meant that Member State legislatures could not stipulate that a specified kind of electronic signature must be used in particular circumstances.

Indeed Recital (49) of eIDAS makes clear that the only limitation on Member States’ ability to define the legal effect of an electronic signature is the stipulation for handwritten signature equivalence provided by Article 25.2:
“It is for national law to define the legal effect of electronic signatures, except for the requirements provided for in this Regulation according to which a qualified electronic signature should have the equivalent legal effect of a handwritten signature.”

The Law Commission Report on Electronic Execution of Documents observed at para 3.15:
“eIDAS therefore allows member states to make provision for the legal effect of electronic signatures which are not qualified electronic signatures. This would allow member states to lay down, for example, security standards to be complied with by e-signing systems should they want to.”

Furthermore, Article 2(3) of eIDAS reserves to national or other EU law the imposition of formalities.  The Regulation:
“does not affect national or Union law related to the conclusion and validity of contracts or other legal or procedural obligations relating to form”

Recital (21) reinforces the point:
“Neither should this Regulation cover aspects related to the conclusion and validity of contracts or other legal obligations where there are requirements as regards form laid down by national or Union law. In addition, it should not affect national form requirements pertaining to public registers, in particular commercial and land registers.”

Obligations relating to form may include formalities relating to signatures. The Law Commission took the view (at para 3.34) that such formalities could include a requirement for witnessing of a signature, or that the signature be in a specific form (such as handwritten).

eIDAS therefore does not prevent a Member State from enacting legislation that stipulates formalities that have to be complied with for a particular purpose, including use of a particular kind of signature (whether wet-ink or electronic).

Compliance with a statutory requirement for a signature, potentially affecting the validity of the underlying document or transaction, is best understood as a separate matter from ‘legal effect’ of a signature. As such it is a matter of domestic English law, untouched by eIDAS.

If ‘shall not be denied legal effect’ in Article 25(1) does not refer to satisfying a statutory requirement for a signature, what does it mean? The answer probably lies in the uses to which signatures are generally put in the absence of a statutory requirement. As already discussed, if a person adopts the contents of a document by means of their signature, any legal effect will vary depending on the kind of document and the legal issue that has subsequently arisen.

The European Commission, in its eIDAS Questions and Answers, says:

            “What do the eIDAS non-discrimination clauses mean?

The eIDAS Regulation sets the principle of non-discrimination of the legal effects and admissibility of electronic signatures … as evidence in legal proceedings. Courts (or other bodies in charge of legal proceedings) cannot discard them as evidence only because they are in an electronic form. Nevertheless, Courts must check whether there are any procedures to be followed according to the EU or national (general or sectorial) law for a given document (including possible requirements on the use of specific levels of electronic tools) and might discard them on these grounds. In other words, the non-discrimination clause does not mean that each and every procedure can be carried out electronically. It means that Courts have to assess these electronic tools in the same way they would do for their paper equivalent.”

This suggests a limited application of the non-discrimination principle, interpreting ‘legal effect’ as requiring only non-discriminatory application of national court procedures. On this basis only a bright line categorical refusal by a court to consider electronic signatures as a class would be impermissible.

Also, Recital (49) of eIDAS (above) suggests that Member States may stipulate the legal effect of electronic signatures (other than qualified signatures), even if that legal effect is specific to a particular kind of electronic signature.

Article 25(1) may thus mean that a court considering an electronic signature used in a non-mandated context cannot categorically preclude it from having any legal effect simply because it is electronic; but that Member States may (a) define the legal effect of an electronic signature other than a qualified signature and (b) mandate that particular kinds of signature (electronic or otherwise) must be used for some kinds of document. A court would still be free to deny an electronic signature legal effect on its merits (or lack of them) - for instance on the ground that the particular electronic signature that had been used lacked sufficient probative value. 

That sits well alongside the second limb of Article 25(1), which provides that an electronic signature must be admissible in evidence. Admissibility means only that the court can look at the evidence. The court is then able to consider what evidential weight to give to the signature, for the purpose of evaluating whatever legal significance the signature may have in the context of the dispute on which the court is adjudicating. Article 25(1) does not prescribe that any particular evidential weight should be given to any kind of signature.

As to admissibility in evidence, from an English law perspective admissibility is trivial. There was never any doubt that an electronic signature is admissible in evidence in an English court. For good measure, that was made explicit in the Electronic Communications Act 2000, which implemented the Electronic Signatures Directive.

Riddle 2: equivalent legal effect to a handwritten signature

Article 25(2) provides that a qualified electronic signature shall have the equivalent legal effect of a handwritten signature. The implicit premise of Article 25.2 is that a handwritten signature has some particular (presumably greater) legal effect than some other kinds of signature. eIDAS does not say what it means by a handwritten signature.

Whether any distinction between handwritten and other signatures exists is a matter of the underlying law of each Member State. Recital (20) of the Directive recognised that: “national law lays down different requirements for the legal validity of handwritten signatures”. Equally, Member States may lay down different legal effects for handwritten signatures recognised as valid under their laws.

To an English lawyer an assumed distinction between handwritten signatures and others is conceptually puzzling, since (as discussed above) handwritten signatures generally have no special legal status in English law. A signature is a signature, whether it be a flowery autograph inscribed using a fountain pen, a rubber stamp facsimile, or an X marked with a pencil. What matters is whether the putative signatory applied it with intent to sign the document[2].

Given the underlying variety of Member State physical signature laws, Article 25(2) can be understood to mean that if under a Member State’s law a handwritten signature as such has some particular legal effect, then a qualified signature must be accorded equivalent legal effect.

Conversely, however, Article 25(2) does not say that only a qualified signature can be accorded equivalent legal effect to a handwritten signature. It is therefore open to a Member State to treat electronic signatures generally as having equivalent legal effect to handwritten and other kinds of physical signature, thus bypassing the potential difficulty of being required to accord an assumed but non-existent special status to a qualified signature.

That is the position that has been adopted in England: any kind of electronic signature is capable of performing the function of a signature.

When we consider the millions of informal electronic signatures used every day, one shudders to think of the havoc that would have been wrought had the Directive (and now eIDAS) stipulated that a Member State could confer legal effect equivalent to a handwritten signature only on a qualified signature. Fortunately, that is not what it says.

Riddle 3: uniquely linked

“Advanced electronic signature” is the silver standard defined under eIDAS. Unlike for the gold standard “qualified electronic signature”, eIDAS confers no particular legal status on an advanced electronic signature. It is intended as a defined category of signature that can be referred to in other EU or Member State legislation or in private documents such as contracts.  But it is also a component of the qualified signature which, as we have seen, must be accorded equivalent legal effect to a handwritten signature.

The definition of an advanced electronic signature sets another puzzle. An advanced signature must satisfy four conditions:

  • The signature is uniquely linked to the signatory
  • It is capable of identifying the signatory
  • It is created using electronic signature creation data that the signatory can, with a high level of confidence, use under his sole control. “electronic signature creation data” means unique data which is used by the signatory to create an electronic signature. An “electronic signature” is data in electronic form which is attached to or logically associated with other data in electronic form and which is used by the signatory to sign.
  • It is linked to the signed data in such a way that any subsequent change in the data is detectable.
The first condition is that the signature is uniquely linked to the signatory. But what does that mean? Is the required link technical, logical or some other kind of association?

Since a signatory (“a natural person who creates an electronic signature”) has to be a human being, how can data be linked in any technical sense to a human being? It is possible to link data to a device, or to other data. But the device is not the human being. What kind of technical feature would be capable of linking an electronic signature to a human being, whether or not uniquely?[3]

Does ‘uniquely linked’ perhaps imply some kind of assurance that the signatory is who s/he says s/he is? If so, that does not go so far as to require third party verification by certification authority. Nothing in eIDAS suggests that a third party certificate was intended to be required by the definition of advanced signature.

A middle ground interpretation could be that if some kind of remembered or biometric information, such as a password or a fingerprint, is required in order to use the unique signature creation data, or perhaps to log on to the machine on which the signing facility is installed, that creates a sufficient logical link between the signature and the signatory[4].

Another possibility might be that it is sufficient if the signature technology is based on data that is unique to the signatory. In this interpretation, “uniquely linked” implies unique association with the signatory, but not necessarily a technological link such as a password to make use of the unique data. However this interpretation seems unlikely, since uniqueness of the data is already a requirement of the third condition.

Fortunately, this riddle rarely requires a solution in the English law context, since there are few examples of legislation that refer to a bare advanced signature. However, some do exist. For instance Reg 37(7) of the Risk Transformation Regulations 2017 provides that where the Financial Conduct Authority certifies an electronic copy as a “true copy”, it must do so with an advanced electronic signature.

The relevance of eIDAS to English signatures law

We have suggested that eIDAS, for all its complexity and technical sophistication, sits on the sidelines as far as the ability to use electronic signatures under English law is concerned. We can now summarise the reasons why this is so:

  • No English statute requires use of an eIDAS-compliant qualified signature
  • Very few English statutes require use of an eIDAS-compliant advanced signature
  • English law already contained no bar, either substantive or in terms of admissibility in evidence, against a signature as such being in electronic form
  • eIDAS does not preclude a national law stipulation for a particular kind of signature. In any event most English law statutory signature requirements are stated in general terms (e.g. ‘signed by’, or ‘signed by or on behalf of’), which under the English law of signatures is capable of being satisfied by any kind of electronic signature.
  • Since English law confers no special status on a handwritten signature, the eIDAS requirement to give equivalent effect to a qualified signature is redundant. In any event English law gives effect to any kind of electronic signature, whether a qualified signature or not.
  • Most impediments to use of electronic signatures under English law are caused not by signature rules, but other formalities governing medium, process or form (such as witnessing). These are outside the scope of eIDAS.
As a matter of practice, within this liberal framework parties deploy electronic signatures by the million, choosing anything from an informal name typed at the end of an e-mail, to signature buttons, to the varieties of signature offered by signing platforms — all according to the nature, value and significance of the transaction. Whether or not to use an eIDAS-defined advanced or qualified signature is a matter of choice on their merits.

In the UK (and indeed in most EU countries) the use of qualified signatures pre-eIDAS was minimal. A 2012 European Commission Staff Working Document recorded one UK qualified certificate provider, which had issued one certificate. As for advanced signatures, under the Directive an advanced signature was generally thought to require the use of a physical signing device such as a smart card, which for signature applications were barely used in the UK.

Why, if a legal system can cope flexibly with all kinds of electronic signature, did EU digital signatures law go down its standards-based path and end up with the eIDAS ecosystem?

For at least part of the answer, we have to delve back into the history of the predecessor Electronic Signatures Directive.

History of the Directive

The Directive was hatched in the late 1990s. The initial focus of the project was on cryptographic digital signatures rather than on electronic signatures generally. This was at a time when it was widely asserted that addressing the perceived electronic non-repudiation problem required the use of cryptographically assured public key-private key digital signatures. Italy, a country traditionally wedded to formalities, had promulgated a highly prescriptive digital signatures decree. Germany was also heading down a technology-specific PKI path.

This was a classic situation in which the Commission could see that national legislation would erect technical barriers within the Internal Market. Hence the enthusiasm to head that off with EU-wide legislation.

The Directive had its origins in the Commission’s 1997 Communication “Ensuring Security and Trust in Electronic Communication”. This aimed to carve out use of encryption for digital signatures from the broader ‘crypto-wars’ that were raging at the time. As the Communication put it: “discussions about the possible conflict between divergent interests on security” had shown “a considerable amount of confrontation and discontent between institutions and interest groups”.

The Communication focused on how to promote use of cryptographic digital signatures as the solution to lack of security and trust, which were said to be an impediment to electronic commerce.

There was also a characteristic European Commission aim of “stimulating a European industry for cryptographic services and products”. The Commission observed that only a few companies in Europe had so far taken steps to offer digital signature services. It averred: “One of the main reasons is the weakness of demand resulting partly from the absence of legal recognition of digital signatures”.

After asserting that important documents could not be exchanged across open networks because of the absence of contractual and mutual trust arrangements present in closed networks, the Commission suggested that ‘authentication and integrity services are needed for secure and trustworthy data transmission and communication over open networks”.

The Commission was particularly focused on cryptographic digital signatures supported by certificates issued by trusted third party Certification Authorities: “In particular CAs are crucial for digital signatures to become a fully accepted tool within national legal systems, for instance to ensure legal recognition and enforceability of a signature in electronic commerce.” 

But the underlying premise of all this — that technically sophisticated digital signatures were a river waiting to flood once EU legislation broke the dam — was little more than an assumption.

So, for example, when the Communication observed that “a key used to authorise a large financial transfer between two banks will require a high level of trust whilst one used to validate a low value personal purchase will not need to be trusted to the same extent”, it did not go on to question why, for the low value personal purchase, anyone would go to the trouble of employing a key-based digital signature or a validation mechanism at all.

A less politically driven project might have placed more emphasis on testing whether assumptions about the degree of trust needed from an electronic signature reflected reality.

By the time of its May 1998 proposal for the Directive the Commission had backed off a little from its focus on PKI and decided that it had to take a more technology-neutral approach:
“Since a variety of authentication mechanisms is expected to develop, the scope of this Directive should be broad enough to cover a spectrum of “electronic signatures”, which would include digital signatures based on public-key cryptography as well as another means of authenticating data”.

Against this background the advanced electronic signature (which as regards ‘uniquely linked’ was defined in the same terms in the Directive as in eIDAS) can be understood as an attempt to describe, in abstract terms, the features of a public key-private key digital signature:

  • The certificate contains data identifying the signatory.
  • The private key is unique.
  • The signature data is technically linked to the private key.
  • The hash function renders the signed document tamper-evident.
However, the Commission’s approach still seemed to assume that an electronic signature of any kind would need some kind of authentication mechanism, whether PKI or some future technology:
“The proposal for a Directive aims at “enabling” the use of electronic signatures within an area without internal frontiers by focusing on the essential requirements for certification services…” [p.5]

The Proposal still envisaged electronic signatures as the kind of thing that would involve internationally agreed standards to establish “an open environment for interoperable products and services” [p.3 pt 6]

Recital (6) of the Proposal for Directive, in a passage that did not make it into the Directive itself, observed that “digital signatures based on public-key cryptography are currently the most recognised form of electronic signature”.

Thus the focus was — perhaps unsurprisingly in an Internal Market Directive — on the technically advanced kinds of signatures that it was hoped would stimulate a future cryptographic services industry within the EU. 

Of course typing a name, or pasting a scan of a manuscript signature, into a document have no authentication mechanism beyond inclusion in the document intended to be signed. They need no service industry infrastructure to support them. In English law, as we have seen, both are capable of functioning as a signature, even where a signature is required by a statute. 

Although not firmly established at that time, that was even then a reasonable deduction from previous English law relating to physical signatures. Only a few years later, in 2001, the English Law Commission opined that:
“Digital signatures, scanned manuscript signatures, typing one’s name (or initials) and clicking on a website button are, in our view, all methods of signature which are generally capable of satisfying a statutory signature requirement. We say that on the basis that it is function, rather than form, which is determinative of the validity of a signature. These methods are all capable of satisfying the principal function: namely, demonstrating an authenticating intention.”

Against the background of the 1997 Communication and the 1998 Proposal, and the emphasis on a hoped-for future digital signatures industry, the subsequent inclusion of broadly defined ordinary electronic signatures in the Directive (defined to include any kind of signature in electronic form) has the impression of being something of an afterthought.

But for English law at any rate, that — and the ability to provide equivalence to a handwritten signature for any kind of electronic signature — were highly significant.  It meant that the Directive changed little or nothing, since as we have seen the common law was already flexible in its approach to what constituted a signature.

In the event, takeup of qualified signatures in the years following the Directive was modest, and in the UK almost non-existent. People and businesses tended to use whatever kind of electronic signature suited their purpose best – even down to the most informal, such as typing a name at the end of an e-mail or into a web form.

Cloud-based signing platforms eventually became popular, but for the most part offered ‘good enough’ signature methods that did not seek to conform to the Directive’s advanced and qualified signature standards. For most ordinary purposes that sufficed and, in the absence of a statutory requirement for an advanced or qualified signature, no-one in England had much reason to worry about whether a signature conformed to any of the Directive’s specifications.

Revising the Directive

When it came to the revision of the Directive, the Commission determined that the modest take-up of standards-based signatures was largely due to lack of cross-border recognition within the EU.

The Commission’s 2012 Proposal for a Regulation displayed the same equation of electronic signatures and sophisticated technology that had been apparent in the 1990s. 

The Staff Working Paper that accompanied the Proposal suggested that a reason for modest take up of electronic signatures was that signing a document or email was “not handy”, that to install a certificate on the computer was “uneasy” and that most applications for private use badly integrated e-signature functionalities.

It went on to say that “free webmail services (such as Hotmail, Yahoo or Gmail) do not allow signing e-mails”. The notion that a user could validly sign an email by typing their name at the foot of it was absent. With this underlying mindset, it is perhaps no surprise that eIDAS turned out to be largely orthogonal to English signatures law.

For signatures, the eIDAS Regulation made two main changes compared with the Directive: it made clear that the ‘gold standard’ signature (now called a ‘qualified signature’) could be implemented remotely in the cloud, not just by a physical device such as a USB signature dongle or a smart card; and it introduced a system of intra-EU cross-border recognition of trust service providers (who provide the third party certificates that underpin qualified signatures).

With these eIDAS changes, providers are now offering standards-based advanced and qualified signatures. Whether there will be an increase in the appetite to use them will no doubt become apparent in time.

[1] Lorna Brazell, in Electronic Signatures and Identities Law and Regulation (3rd ed Sweet & Maxwell, 2018, para 2-002) identifies as many as seven potential functions of a signature.
[2] However, other questions may arise such as whether the signatory can delegate the act of signing to an agent.
[3] Stephen Mason has suggested that the ‘uniquely linked’ condition is impossible to comply with. See Electronic Signatures in Law (4th ed, 2016), para 4.17.
[4] Cf. Brazell (op cit) at 6-054: “some unique logical link”, such as “independent evidence as to who it was who applied the signature means”.