Ofcom’s proactive technology measures: principles-based or vague?

Ofcom has published its long-expected consultation on additional measures that it recommends U2U platforms and search engines should implement to fulfil their duties under the Online Safety Act.  The focus, this time, is almost entirely on proactive technology: automated systems intended to detect particular kinds of illegal content and content harmful to children, with a view to blocking or swiftly removing them.

The consultation marks a further step along the UK’s diverging path from the EU Digital Services Act. The DSA prohibits the imposition of general monitoring obligations on platforms. Those are just the kind of obligations envisaged by the Online Safety Act’s preventative duties, which Ofcom is gradually fleshing out and implementing.

Ofcom finalised its first Illegal Harms Code of Practice in December 2024. For U2U services the Code contained two proactive technology recommendations: hash and URL matching for CSAM. The initial consultation had also suggested fuzzy keyword matching to detect some kinds of fraud, but Ofcom did not proceed with that. The regulator indicated that it would revisit fraud detection in a later, broader consultation. That has now arrived.

The new U2U proposals go beyond fraud. They propose perceptual hash-matching for visual terrorism content and for intimate image abuse content. They suggest that content should be excluded from recommender feeds if there are indications that it is potentially illegal, unless and until it is determined via content moderation to be legal. 

Most ambitiously, Ofcom wants certain relatively large platforms to research the availability and suitability (in accordance with proposed criteria) of proactive technology for detection of fraud and some other illegal behaviour, then implement it if appropriate. Those platforms would also have to review existing technologies that they use for these purposes and, if feasible, bring them into line with Ofcom’s criteria.

Ofcom calls this a ‘principles-based’ measure, probably because it describes a qualitative evaluation and configuration process rather than prescribing any concrete parameters within which the technology should operate.

Freedom of expression

Legal obligations for proactive content detection, blocking and removal engage the fundamental freedom of expression rights of users. Obligations must therefore comply with ECHR human rights law, including requirements of clarity and certainty.

Whilst a principles-based regime may be permissible, it must nevertheless be capable of predictable application. Otherwise it will stray into impermissible vagueness. Lord Sumption in Catt said that what is required is a regime the application of which is:

“reasonably predictable, if necessary with the assistance of expert advice. But except perhaps in the simplest cases, this does not mean that the law has to codify the answers to every possible issue which may arise. It is enough that it lays down principles which are capable of being predictably applied to any situation."

In Re Gallagher he said that:

“A measure is not “in accordance with the law” if it purports to authorise an exercise of power unconstrained by law. The measure must not therefore confer a discretion so broad that its scope is in practice dependent on the will of those who apply it, rather than on the law itself. Nor should it be couched in terms so vague or so general as to produce substantially the same effect in practice.”

Typically these strictures would apply to powers and duties of public officials. The Online Safety Act is different: it requires U2U service providers to make content decisions and act (or not) to block or remove users’ posts. Thus the legal regime that requires them to do that has to provide sufficient predictability of their potential decisions and resulting acts.

In addition to fraud and financial services offences, Ofcom’s proposed principles-based measures would apply to image based CSAM, CSAM URLs, grooming, and encouraging or assisting suicide (or attempted suicide).

Any real-time automated content moderation measure poses questions about human rights compatibility. The auguries are not promising: proactive technology, armed only with the user’s post and perhaps some other on-platform data, will always lack contextual information. For many offences off-platform information can be the difference between guilt and innocence.  Decisions based on insufficient information inevitably stray into arbitrariness.

Then there is the trade-off between precision and recall. Typically, the more target content the automated tool is tuned to catch, the more false positives it will also throw up. False positives result in collateral damage to legitimate speech. It does not take many false positives to constitute disproportionate interference with users’ rights of freedom of expression.

Lord Grade, the Chairman of Ofcom, said in a recent speech that the aims of tackling criminal material and content that poses serious risks of harm to children’s physical or emotional health were not in conflict with freedom of expression. Indeed so, but focusing only on the aim misses the point: however worthy the end, it is the means - in this case proactive technology - that matters.

Prescribed by law

Ofcom’s Proactive Technology Draft Guidance says this about proportionality of the proposed measures:

“Proactive technology used for detection of harmful content involves making trade-offs between false positives and false negatives. Understanding and managing those trade-offs is essential to ensure the proactive technology performs proportionately, balancing the risk of over-removal of legitimate content with failure to effectively detect harm.” (para 5.14)

Proportionality is a requirement of human rights compliance. However, before considering proportionality a threshold step has to be surmounted: the ‘prescribed by law’ or ‘legality’ condition. This is a safeguard against arbitrary restrictions - laws should be sufficiently precise and certain that they have the quality of law.

The prescribed by law requirement is an aspect of the European Convention on Human Rights. It has also been said to be a UK constitutional principle that underpins the rule of law:

"The acceptance of the rule of law as a constitutional principle requires that a citizen, before committing himself to any course of action, should be able to know in advance what are the legal consequences that will flow from it." (Lord Diplock, Black-Clawson [1975])

The Constitutional Reform Act 2005 refers in S.1 to:

“the existing constitutional principle of the rule of law”.

For content monitoring obligations the quality of law has two facets, reflecting the potential impact of the obligations on the fundamental rights of both platforms and users.

The platform aspect is written in to the Act itself:

“the measures described in the code of practice must be sufficiently clear, and at a sufficiently detailed level, that providers understand what those measures entail in practice”. (Schedule 4)

The user aspect is not spelled out in the Act but is no less significant for that. Where a user’s freedom of speech may be affected by steps that a platform takes to comply with its duties, any interference with the user’s right of freedom of expression must be founded on a clear and precise rule.

That means that a user must be able to foresee in advance with reasonable certainty whether something that they have in mind to post is or is not liable to be blocked, removed or otherwise affected as a result of the obligations that the Act places on the platform.

That is not simply a matter of users themselves taking care to comply with substantive law when they consider posting content. The Act interpolates platforms into the process and may require them to make judgements about whether the user’s post is or is not illegal. Foreseeability is therefore a function both of the substantive law and of the rules about how a platform should make those judgements.

If, therefore, the mechanism set up by the Act and Ofcom for platforms to evaluate, block and take down illegal content is likely to result in unpredictable, arbitrary determinations of what is and is not illegal, then the mechanism fails the ‘prescribed by law’ test and is a per se violation of the right of freedom of expression.

Equally, if the regime is so unclear about how it would operate in practice that a court is not in a position to assess its proportionality, that would also fail the ‘prescribed by law’ test. That is the import of Lord Sumption’s observations in Catt and Gallagher (above).

A prescriptive bright-line rule, however disproportionate it might be, would satisfy the ‘prescribed by law’ test and fall to be assessed only by reference to necessity and proportionality. Ofcom’s principles-based recommendations, however, are at the opposite end of the spectrum: they are anything but a bright-line rule. The initial ‘prescribed by law’ test therefore comes into play.

How do Ofcom’s proposed measures stack up?

Service providers themselves would decide how accurate the technology has to be, what proportion of content detected by the technology should be subjected to human review, and what is an acceptable level of false positives.

Whilst Ofcom specifies various ‘proactive technology criteria’, those are expressed as qualitative factors to be taken into account, not quantitative parameters. Ofcom does not specify what might be an appropriate balance between precision and recall, nor what is an appropriate proportion of human review of detected content.

Nor does Ofcom indicate what level of false positives might be so high as to render the technology (alone, or in combination with associated procedures) insufficiently accurate.

Examples of Ofcom’s approach include:

“However, there are some limitations to the use of proactive technology in detecting or supporting the detection of the relevant harms. For example, proactive technology does not always deal well with nuance and context in the same way as humans.

However, we mitigate this through the proactive technology criteria which are designed to ensure proactive technology is deployed in a way that ensures an appropriate balance between precision and recall, and that an appropriate proportion of content is reviewed by humans.” (Consultation, para 9.92)

“Where a service has a higher tolerance for false positives, more content may be wrongly identified. … The extent of false positives will depend on the service in question and the way in which it configures its proactive technology. The measure allows providers flexibility in this regard, including as to the balance between precision and recall (subject to certain factors set out earlier in this chapter).” (Consultation, paras 9.135, 9.136)

“… when determining what is an appropriate proportion of detected content to review by humans, providers have flexibility to decide what proportion of detected content it is appropriate to review, however in so doing, providers should ensure that the following matters are taken into account…” (Consultation, para 9.19)

“However, in circumstances where false positives are consistently high and cannot be meaningfully reduced or mitigated, particularly where this may have a significant adverse impact on user rights, providers may conclude that the proactive technology is incapable of meeting the criteria.” (Proactive Technology Draft Guidance, para 5.19)

How high is high? How significant is significant? No answer is given, other than that the permissible level of false positives is related to the nature of the subsequent review of detected content. As we shall see, the second stage review does not require all content detected by the proactive technology to be reviewed by human beings. The review could, seemingly, be conducted by a second automated system.

The result is that two service providers in similar circumstances could arrive at completely different conclusions as to what constitutes an acceptable level of legitimate speech being blocked or taken down. Ofcom acknowledges that the flexibility of its scheme:

“could lead to significant variation in impact on users’ freedom of expression between services”. (Consultation, para 9.136)

That must raise questions about the predictability and foreseeability of the regime.

If the impact on users’ expression is not reasonably foreseeable, that is a quality of law failure and no further analysis is required. If that hurdle were surmounted, there is still the matter of what level of erroneous blocking and removal would amount to a disproportionate level of interference with users’ legitimate freedom of expression. 

Proportionality?

Ofcom concludes that:

“Having taken account of the nature and severity of the harms in question, the principles we have built into the measure to ensure that the technology used is sufficiently accurate, effective and lacking in bias, and the wider range of safeguards provided by other measures, we consider overall that the measure’s potential interference to users’ freedom of expression to be proportionate.” (Consultation, para 9.154)

However, it is difficult to see how Ofcom (or anyone else) can come to any conclusion as to the overall proportionality of the recommended principles-based measures when they set no quantitative or concrete parameters for precision versus recall, accuracy of review of suspect content, or an ultimately acceptable level of false positives.

Ofcom’s discussion of human rights compliance starts with proportionality. While it notes that the interference must be ‘lawful’, there is no substantive discussion of the ‘prescribed by law’ threshold.

Prior restraint

Finally, on the matter of human rights compatibility, proactive detection and filtering obligations constitute a species of prior restraint (Yildirim v Turkey (ECtHR), Poland v The European Parliament and Council (CJEU)).

Prior restraint is not impermissible. However, it does require the most stringent scrutiny and circumscription, in which risk of removal of legal content will loom large. The ECtHR in Yildirim noted that “the dangers inherent in prior restraints are such that they call for the most careful scrutiny on the part of the Court”.

The proactive technology criteria

Ofcom’s proactive technology criteria are, in reality, framed not as a set of criteria but as a series of factors that the platform should take into account.  Ofcom describes them as “a practical, outcomes-focused set of criteria.” [Consultation, para 9.13]

Precision and recall One criterion is that the technology has been evaluated using “appropriate” performance metrics and

“configured so that its performance strikes an appropriate balance between precision and recall”.  (Recommendation C11.3(c))

Ofcom evidently must have appreciated that, without elaboration, “appropriate” was an impermissibly vague determinant. The draft Code of Practice goes on (Recommendation C11.4(a)):

“when configuring the technology so that it strikes an appropriate balance between precision and recall, the provider should ensure that the following matters are taken into account:

i) the service’s risk of relevant harm(s), reflecting the risk assessment of the service and any information reasonably available to the provider about the prevalence of target illegal content on the service;

ii) the proportion of detected content that is a false positive;

iii) the effectiveness of the systems and/or processes used to identify false positives; and

iv) in connection with CSAM or grooming, the importance of minimising the reporting of false positives to the National Crime Agency (NCA) or a foreign agency;”

These factors may help a service provider tick the compliance boxes – ‘Yes, I have taken these factors into account’ - but they do not amount to a concrete determinant of what constitutes an appropriate balance between precision and recall.

Review of detected content Accuracy of the proactive technology is, as already alluded to, only the first stage of the recommended process. The service provider has to treat a detected item as providing ‘reason to suspect’ that it is illegal content, then move on to a second stage: review.

“Where proactive technology detects or supports the detection of illegal content and/or content harmful to children, providers should treat this as reason to suspect that the content may be target illegal content and/or content harmful to children.

Providers should therefore take appropriate action in line with existing content moderation measures, namely ICU C1 and ICU C2 (in the Illegal Content User-to-user Codes of Practice) and PCU C1 and PCU C2 (in the Protection of Children User-to-user Code of Practice), as applicable.” (Consultation, para 9.74)

That is reflected in draft Codes of Practice paras ICU C11.11, 12.9 and PCU C9.9, 10.7. For example:

“ICU C11.11 Where proactive technology detects, or supports the detection of, target illegal content in accordance with ICU C11.8(a), the provider should treat this as reason to suspect that the content may be illegal content and review the content in accordance with Recommendation ICU C1.3.”

‘Review’ does not necessarily mean human review. Compliance with the proactive technology criteria requires that:

“...policies and processes are in place for human review and action is taken in accordance with that policy, including the evaluation of outputs during development (where applicable), and the human review of an appropriate proportion of the outputs of the proactive technology during deployment. Outputs should be explainable to the extent necessary to support meaningful human judgement and accountability.” (Emphasis added) (draft Code of Practice Recommendation ICU C11.3(g))

The consultation document says:

“It should be noted that this measure does not itself recommend the removal of detected content. Rather, it recommends that providers moderate detected content in accordance with existing content moderation measures (subject to human review of an appropriate proportion of detected content, as mentioned above).” (Consultation, para 9.147)

And:

“Providers have flexibility in deciding what proportion of detected content is appropriate to review, taking into account [specified factors]” (Consultation, para 9.145)

Ofcom has evidently recognised that “appropriate proportion” is, without elaboration, another impermissibly vague determinant. It adds (Recommendation C11.4(b)):

“when determining what is an appropriate proportion of detected content to review by humans, the provider should ensure that the following matters are taken into account:

i) the principle that the resource dedicated to review of detected content should be proportionate to the degree of accuracy achieved by the technology and any associated systems and processes;

ii) the principle that content with a higher likelihood of being a false positive should be prioritised for review; and

iii) in the case of CSAM or grooming, the importance of minimising the reporting of false positives to the NCA or a foreign agency.”

As with precision and recall, these factors may help a service provider tick the compliance boxes but are not a concrete determinant of the proportion of detected content to be submitted to human review in any particular case.

Second stage review – human, more technology or neither?

The upshot of all this appears to be that content detected by the proactive technology should be subject to review in accordance with the Code of Practice moderation recommendations; and that an ‘appropriate proportion’ of that content should be subject to human review.

But if only an ‘appropriate proportion’ of content detected by the proactive technology has to be subject to human review, how is the rest to be treated? Since it appears that some kind of ‘appropriate action’ is contemplated in accordance with Ofcom’s content moderation recommendations, the implication appears to be that moderation at the second stage could be by some kind of automated system.

In that event it would seem that the illegal content judgement itself would be made by that second stage technology in accordance with Recommendation C1.3.

Recommendation C1.3, however, does not stipulate the accuracy of second stage automated technology. The closest that the Code of Practice comes is ICU C4.2 and 4.3:

“The provider should set and record performance targets for its content moderation function, covering at least:

a) the time period for taking relevant content moderation action; and

b) the accuracy of decision making.

In setting its targets, the provider should balance the need to take relevant content moderation action swiftly against the importance of making accurate moderation decisions.”

Once again, the path appears to lead to an unpredictable balancing exercise by a service provider.

Curiously, elsewhere Ofcom appears to suggest that second stage “complementary tools” could in some cases merely be an ‘additional safeguard’:

“What constitutes an appropriate balance between precision and recall will depend on the nature of the relevant harm, the level of risk identified and the service context. For example, in some cases a provider might optimise for recall to maximise the quantity of content detected and apply additional safeguards, such as use of complementary tools or increased levels of human review, to address false positives. In other cases, higher precision may be more appropriate, for example, to reduce the risk of adverse impacts on user rights.” (Proactive Technology Draft Guidance, para 5.18)

If the implication of ‘in some cases’ is that in other cases acting on the output of the proactive technology without a second stage review would suffice, that would seem to be inconsistent with the requirement that all detected content be subject to some kind of moderation in accordance with Recommendation C1.3.

Moreover, under Ofcom’s scheme proactive technology is intended only to provide ‘reason to suspect’ illegality. That would not conform to the standard stipulated by the Act for an illegal content judgement: ‘reasonable grounds to infer’.

Conclusion

When, as Ofcom recognises, the impact on users’ freedom of expression will inevitably vary significantly between services, and Ofcom’s documents do not condescend to what is or is not an acceptable degree of interference with legitimate speech, it is difficult to see how a user could predict, with reasonable certainty, how their posts are liable to be affected by platforms’ use of proactive technology in compliance with Ofcom’s principles-based recommendations.

Nor is it easy to see how a court would be capable of assessing the proportionality of the measures. As Lord Sumption observed, the regime should not be couched in terms so vague or so general as, substantially, to confer a discretion so broad that its scope is in practice dependent on the will of those who apply it. Again, Ofcom's acknowledgment that the flexibility of its scheme could lead to significant variation in impact on users’ freedom of expression does not sit easily with that requirement.  

Ofcom, it should be acknowledged, is to an extent caught between a rock and a hard place. It has to avoid being overly technology-prescriptive, while simultaneously ensuring that the effects of its recommendations are reasonably foreseeable to users and capable of being assessed for proportionality. Like much else in the Act, that may in reality be an impossible circle to square. That does not bode well for the Act’s human rights compatibility.

[Amended 6 August 2025 to add ‘principles-based’ to the first paragraph of the Conclusion.]

The Ordinary Reasonable Person encounters (or not) cyber-abuse

The recent decision of the Australian Administrative Review Tribunal in X Corp and Elston v eSafety Commissioner illustrates the complexities that can arise when the law tasks a regulator or platform to adjudge an online post.

The decision grapples with a dilemma that is familiar, albeit under a very different legislative regime, from the UK’s Online Safety Act 2023. It is also features in the police takedown notice scheme for unlawful knives and other weapons content contained in the Crime and Policing Bill (currently making its way through Parliament).

At a high level, the issue is how to achieve rapid removal of impugned user content (typically because it is illegal under the general law or defined as harmful in some way), while not affecting legitimate posts. The specific challenge is that the contents of the post alone are often insufficient to determine whether the legal line has been crossed. Contextual information, which may be off-platform and involve investigation, is required. The Elston case provides a vivid illustration.

The twin imperatives of rapid removal and adequate investigation of context stand in conflict with each other. A regime that requires contravention to be adjudged solely on the contents of a post, ignoring external context, is likely to be either ineffectual or overreaching, depending on which way the adjudicator is required to jump in the absence of relevant information.

Australia’s Online Safety Act 2021 empowers the eSafety Commissioner, but only following receipt of a complaint, to issue a content removal notice to a social media platform if she is satisfied that a user’s post constitutes cyber-abuse material targeted at an Australian adult. (In this respect the Australian legislation resembles the UK Crime and Policing Bill more than our Online Safety Act: Ofcom has no power under the OSA to require removal of a specific item of user content. The Crime and Policing Bill will institute a regime of police takedown notices for unlawful knives and other weapons content, albeit not predicated on receipt of a complaint.)

Cyber-abuse material under the Australian Act has two key elements. The eSafety Commissioner has to be satisfied of both before issuing a removal notice:

Intention Element: an ordinary reasonable person would conclude that it is likely that the material was intended to have an effect of causing serious harm to a particular Australian adult.

Offense Element: an ordinary reasonable person in the position of the Australian adult would regard the material as being, in all the circumstances, menacing, harassing or offensive.

Serious harm is defined as serious physical harm or serious harm to a person’s mental health, whether temporary or permanent. Serious harm to a person’s mental health includes:

(a) serious psychological harm; and

(b) serious distress;

but does not include mere ordinary emotional reactions such as those of only distress, grief, fear or anger.

The need to assess what an ‘ordinary reasonable person’ would think is common to both elements. For the Intention Element the Ordinary Reasonable Person has to determine the likely intention of the person who posted the material. For the Offense Element, in order to determine how the material should be regarded, the Ordinary Reasonable Person has to be put in the position of the Australian adult putatively intended to be targeted.

The reason why the legislation hypothesises an Ordinary Reasonable Person is to inject some objectivity into what could otherwise be an overly subjective test.

The Tribunal observed that the Intention Element converted what would otherwise be “a broadly available censorship tool based on emotional responses to posted material” into a provision that “protects people from a much narrower form of conduct where causing serious harm to a particular person was, in the relevant sense, intended” [21]. (This has similarities to the heavy lifting done by the mental element in broadly drafted terrorism offences.)

We are in familiar legal territory with fictive characters such as the Ordinary Reasonable Person. It is reminiscent of the fleeting appearance of the Person of Ordinary Sensibilities in the draft UK Online Safety Bill.

Nevertheless, as the Tribunal decision illustrates, the attributes of the hypothetical person may need further elucidation. Those characteristics can materially affect the balance between freedom of expression and the protective elements of the legislation in question.

Thus, what is the Ordinary Reasonable Person taken generally to know?  What information can the Ordinary Reasonable Person look at in deciding whether intention to cause serious harm is likely? How likely is likely?

The information available to the Ordinary Reasonable Person

The question of what information can, or should, be taken into account is especially pertinent to legislation that requires moderation decisions to be made that will impinge on freedom of expression. The Tribunal posed the question thus:

“… whether findings on the Intention Element should be made on an impressionistic basis after considering a limited range of material, or whether findings should be made after careful consideration, having regard to any evidence obtained as part of any investigation or review process.” [45]

It found that:

“The history and structure of the provisions suggest that while impressionistic decision-making may be authorised in the first instance, early decisions made on limited information can and should be re-visited both internally and externally as more information becomes available, including as a result of input from the affected end-user.” [45]

That was against the background that:

“…the legislation as passed allows for rapid decision making by the Commissioner to deal with material that appears, on its face, to be within a category that the Act specified could be the subject of a removal notice. However, once action has been taken, the insertion of s 220A confirms that Parliament accepted that there needed to be an opportunity for those affected by the action to have an opportunity to address whether the material was actually within the prohibited statutory category. External review by the Tribunal was provided for with the same end in mind.” [44]

The UK Online Safety Act states that a platform making an illegality judgement should do so on the basis of all relevant information reasonably available to it. Ofcom guidance fleshes out what information is to be regarded as reasonably available.

The UK Crime and Policing Bill says nothing about what information a police officer giving an unlawful weapons content removal notice, or a senior officer reviewing such a notice, should seek out and take into account. Nor does it provide any opportunity for the user whose content is condemned to make representations, or to be notified of the decision.

Generally speaking, the less information that can or should be taken into account, the greater the likelihood of arbitrary decision-making and consequent violation of freedom of expression rights.

In the Elston case three different variations on the Ordinary Reasonable Person were put to the Tribunal. The eSafety Commissioner argued that the Ordinary Reasonable Person should be limited to considering the poster’s profile on X and the material constituting the post. The poster’s subsequent evidence about his intention and motivations was irrelevant to determining whether the Intention Element was satisfied. The same was said to apply to evidence about the poster’s knowledge of the Australian person said to be targeted. (The Tribunal observed that that would mean that even material contained in the complaint that preceded the removal notice would be excluded from consideration.)

As to the general knowledge of the Ordinary Reasonable Person, the eSafety Commissioner argued that (for the purposes of the case before the Tribunal, which concerned a post linking to and commenting on a newspaper article about a transgender person) the Ordinary Reasonable Person would be aware that material on X can bully individuals, would understand that public discourse around sexuality and gender can be polarising as well as emotionally charged; and would understand that calling a transgender man a woman would be to act contrary to that transgender man’s wishes.

X Corp argued that the decisionmaker was entitled to have regard to evidence (including later evidence) concerning immediate context as at the time of the post, but not more. The facts which could be known to the ordinary reasonable person when making their assessment included facts about the subject of the post or the poster, what their relationship was at the time of the post, but not evidence about what happened after.

The significance of the different positions was that on X Corp’s case, later evidence could be taken into account to the effect that the poster did not know, or know of, the person who was the subject of the post until he read the newspaper article. That was not apparent from the post itself or the poster’s profile.

Mr Elston (the poster) argued that a wide range of material could be acquired and treated as available to the ordinary reasonable person when asked to decide whether the material posted ‘was intended to have an effect of causing serious harm’.

On this view of the statutory power, evidence obtained before or after the post, during the course of the investigation and concerning matters that occurred after the post was made, could be treated as available to the Ordinary Reasonable Person when considering the Intention Element.

On this approach, Mr Elston’s own evidence about his intention would be “relevant to consider, but not necessarily conclusive of what an ordinary reasonable person would conclude about his intention.” [62]

The Tribunal agreed with Mr Elston’s approach:

“The existence of the investigative powers available to the Commissioner and the complaint-based nature of the power provide a powerful basis for concluding that the Commissioner and the Tribunal should be feeding all of the available evidence into the assessment of what the ‘ordinary reasonable person’ would conclude was likely before determining whether the Intention Element is satisfied.” [74]

It added:

“The Parliament was concerned to give end-users an opportunity to address claims about their conduct both on internal review and by providing review in the Tribunal. To read the ordinary reasonable person lens as a basis for disregarding evidence submitted by either the complainant or the end-user or discovered by the Commissioner during an investigation is not consistent with the fair, high quality decision-making the Parliament made provision for.” [77]

The Tribunal then spelled out the consequences of the Commissioner’s approach:

“…In many circumstances, including this case, limiting the information that can be considered by the ‘ordinary reasonable person’ to the post and closely related material, results in critical information not being available.” [81]

It went on:

“In this case, there is no evidence in any of the material posted and associated with the post, that the post was ever brought to the attention of Mr Cook [the complainant]. …

That Mr Cook was aware of the post is only discoverable by reference to the complaint submitted to the Commissioner. If a decision maker is restricted to knowing that a post was made to a limited audience, none of whom included Mr Cook, reaching the conclusion that the material was intended to cause serious harm to Mr Cook is going to be difficult. In those circumstances, where there appears to be no evidence to which the decision maker can have regard in order to make a finding that the post came to Mr Cook’s attention, let alone was intended to come to his attention, a decision to issue a removal notice could not be sustained.” [81]

The Tribunal reiterated:

“In many cases, it will be the complaint that provides critical context to allow an ordinary reasonable person to conclude that serious harm was intended.” [81]

The Tribunal concluded that evidence about what happened after the post was posted could be relevant if it shed light on the likely intention of the poster. Similarly, evidence about prior behaviour of third parties in response to certain posts could be relevant, even if it was only discoverable by the regulator using compulsory powers:

“So long as evidence sheds light on the statutory question, then it can and should be considered. It would be inappropriate in advance of a particular factual scenario being presented to the decision-maker to say that there are whole categories of evidence that cannot be considered because the statutory test in all circumstances renders the material irrelevant.” [87]

Nevertheless, that did not mean that the concept of the ‘ordinary and reasonable’ person had no effect:

“It moves the assessment away from a specific factual inquiry concerning the actual thought process of the poster and what effect they intended to achieve by the post. I must undertake a more abstract inquiry about what an independent person (who isn’t me) would think was the poster’s intention having regard to the available evidence. Provided evidence is relevant to that question, then it can and should be considered.” [89]

Whilst specific to the Australian statute and its fictive Ordinary Reasonable Person, this discussion neatly illustrates the point that has repeatedly been made (and often ignored): that platform judgements as to illegality required by the UK Online Safety Act will very often require off-platform contextual information and cannot sensibly be made on the basis of a bare user post and profile.

The point assumes greater significance with real-time proactive automated content moderation – something that Ofcom is proposing to extend – which by its very nature is unlikely to have access to off-platform contextual information.

The discussion also speaks eloquently to the silence of the Crime and Policing Bill on what kind and depth of investigation a police officer should conduct in order to be satisfied as to the presence of unlawful weapons content.

Likelihood of serious harm

The other significant point that the Tribunal had to consider was what the statute meant by ‘likely’ that serious harm was intended. The rival contentions were ‘real chance’ and ‘more probable than not’. The Tribunal held that, in the statutory context, the latter was right. The conclusion is notable for acknowledging the adverse consequences for freedom of expression of adopting a lower standard:

“A finding by the ordinary reasonable person that a person was setting out to cause serious harm to another is a serious, adverse finding with implications for freedom of expression. It is not the kind of finding that should be made when it is only possible that serious harm was intended.” [119]

The standard set by the UK Online Safety Act for making content illegality judgements is “reasonable grounds to infer”. It remains questionable, to say the least, whether that standard is compatible with ECHR Article 10. The Crime and Policing Bill says no more than that the police officer must be ‘satisfied’ that the material is unlawful weapons content.  

The Tribunal’s conclusion

On the facts of the case, the Tribunal concluded that an ordinary reasonable person in the position of the complainant Mr Cook would regard the post as offensive; but that the Intention Element was not satisfied. That depended crucially on the broader contextual evidence:

“Read in isolation, the post looks to be an attempt to wound Mr Cook and upset him and cause him distress, perhaps even serious distress. If an ordinary reasonable person was only aware of the post, then it may be open to find that the poster’s intention was likely to be to cause serious harm to Mr Cook. However, when the broader context is known and understood, it is difficult to read the post as intended to harm Mr Cook, or intended to have others direct criticism towards Mr Cook or designed to facilitate vitriol by spreading personal information about him.” [191]

Amongst the broader context was lack of evidence that the poster intended the post to come to Mr Cook’s attention.

“For the post to do any harm it needed to be read by Mr Cook. While I am satisfied that Mr Elston was indifferent to whether the post did come to Mr Cook’s attention and indifferent to whether or not it distressed him, there is no evidence to support the conclusion that the post was made with the intention of it being brought to Mr Cook’s attention.” [197]

Part of the reasoning behind that conclusion was that Mr Elston’s post did not tag Mr Cook’s user handle, but only that of the World Health Organisation (which had appointed Mr Cook to an advisory panel):

“ It is notable that Mr Elston only included the handle for the WHO in his post and there is nothing in the body of the post that attempts to facilitate the contacting of Mr Cook by Mr Elston’s followers. Mr Cook’s name is not used in the body of the post.” [200]

Overall, the Tribunal concluded:

“When the evidence is considered as a whole I am not satisfied that an ordinary reasonable person would conclude that by making the post Mr Elston intended to cause Mr Cook serious harm. In the absence of any evidence that Mr Elston intended that Mr Cook would receive and read the post, and in light of the broader explanation as to why Mr Elston made the post, I am satisfied that an ordinary reasonable person would not conclude that that it is likely that the post was intended to have an effect of causing serious harm to Mr Cook.” [207]

For present purposes the actual result in the Elston case matters less than the illustration that it provides of what can be involved in making judgements about removal or blocking of posts against a statutory test: whether that evaluation be done by a regulator, a platform discharging a duty imposed by statute or (in the likely future case of unlawful weapons content) the police.

The Online Safety Act grumbles on

Policymakers sometimes comfort themselves that if no-one is completely satisfied, they have probably got it about right. 

On that basis, Ofcom’s implementation of the Online Safety Act’s illegality duties must be near-perfection: the Secretary of State (DSIT) administering a sharp nudge with his draft Statement of Strategic Priorities, while simultaneously under fire for accepting Ofcom’s advice on categorisation of services; volunteer-led community forums threatening to close down in the face of perceived compliance burdens; and many of the Act’s cheerleaders complaining that Ofcom’s implementation has so far served up less substantial fare than they envisaged. 

As of now, an estimated 25,000 UK user-to-user and search providers (plus another 75,000 around the world) are meant to be busily engaged in getting their Illegal Harms risk assessments finished by 16 March. 

Today is Safer Internet Day. So perhaps spare a thought for those who are getting to grips with core and enhanced inputs, puzzling over what amounts to a ‘significant’ number of users, learning that a few risk factors may constitute ‘many’ (footnote 74 to Ofcom’s General Risk Level Table), or wondering whether their service can be ‘low risk’ if they allow users to post hyperlinks.  (Ofcom has determined that hyperlinks are a risk factor for six of the 17 kinds of priority offence designated by the Act: terrorism, CSEA, fraud and financial services, drugs and psychoactive substances, encouraging or assisting suicide and foreign interference offences). 

Grumbles from whichever quarter will come as no great surprise to those (this author included) who have argued from the start that the legislation is an ill-conceived, unworkable mess which was always destined to end in tears. Even so, and making due allowance for the well-nigh impossible task with which Ofcom has been landed, there is an abiding impression that Ofcom’s efforts to flesh out the service provider duties - risk assessment in particular – could have been made easier to understand. 

The original illegal harms consultation drew flak for its sheer bulk: a tad over 1,700 pages. The final round of illegal harms documents is even weightier: over 2,400 pages in all. It is in two parts. The first is a Statement. In accordance with Ofcom’s standing consultation principles, it aims to explain what Ofcom is going to do and why, showing how respondents’ views helped to shape Ofcom’s decisions. That amounts to 1,175 pages, including two summaries. 

The remaining 1,248 pages consist of statutory documents: those that the Act itself requires Ofcom to produce. These are a Register of Risks, Risk Assessment Guidance, Risk Profiles, Record Keeping and Review Guidance, a User to User Illegal Content Code of Practice, a Search Service Illegal Content Code of Practice, Illegal Content Judgements Guidance, Enforcement Guidance, and Guidance on Content Communicated Publicly and Privately. Drafts of the two Codes of Practice were laid before Parliament on 16 December 2024. Ofcom can issue them in final form upon completion of that procedure.

When it comes to ease of understanding, it is tempting to go on at length about the terminological tangles to be found in the documents, particularly around ‘harm’, ‘illegal harm’ and ‘kinds of illegal harm’. But really, what more is worth saying? Ofcom’s documents are, to all intents and purposes, set in stone. Does it help anyone to pen another few thousand words bemoaning opaque language? Other than in giving comfort that they are not alone to those struggling to understand the documents, probably not. Everyone has to get on and make the best of it.

So one illustration will have to suffice. ‘Illegal harm’ is not a term defined or used in the Act. In the original consultation documents Ofcom’s use of ‘illegal harm’ veered back and forth between the underlying offence, the harm caused by an offence, and a general catch-all for the illegality duties; often leaving the reader to guess in which sense it was being used. 

The final documents are improved in some places, but introduce new conundrums in others. One of the most striking examples is paragraph 2.35 and Table 6 of the Risk Assessment Guidance (emphasis added to all quotations below). 

Paragraph 2.35 says: 

“When evaluating the likelihood of a kind of illegal content occurring on your service and the chance of your service being used to commit or facilitate an offence, you should ask yourself the questions set out in Table 6.”

Table 6 is headed: 

“What to consider when assessing the likelihood of illegal content”

The table then switches from ‘illegal content’ to ‘illegal harm’. The first suggested question in the table is whether risk factors indicate that: 

“this kind of illegal harm is likely to occur on your service?” 

‘Illegal harm’ is footnoted with a reference to a definition in the Introduction: 

“the physical or psychological harm which can occur from a user encountering any kind of illegal content…”. 

So what is the reader supposed to be evaluating: the likelihood of occurrence of illegal content, or the likelihood of physical or psychological harm arising from such content? 

If ‘Illegal Harm’ had been nothing more than a title that Ofcom gave to its illegality workstream, then what the term actually meant might not have mattered very much. But the various duties that the Act places on service providers, and even Ofcom’s own duties, rest on carefully crafted distinctions between illegal content, underlying criminal offences and harm (meaning physical or psychological harm) arising from such illegality. 

That can be seen in this visualisation. It illustrates the U2U service provider illegality duties - both risk assessment and substantive - together with the Ofcom duty to prepare an illegality Risks Register and Risk Profiles.  The visualisation divides the duties into four zones (A, B, C and D), explained below. 

A: The duties in this zone require U2U providers to assess certain risks related to illegal content (priority and non-priority). These risks are independent of and unrelated to harm. The risks to be assessed have no direct counterpart in any of the substantive safety duties in Section 10. Their relevance to those safety duties probably lies in the proportionality assessment of measures to fulfil the Section 10 duties. 

Although the service provider’s risk assessment has to take account of the Ofcom Risk Profile that relates to its particular kind of service, Ofcom’s Risk Profiles are narrower in scope than the service provider risk assessment. Under the Act Ofcom’s Risks Register and Risk Profiles are limited to the risk of harm (meaning physical or psychological harm) to individuals in the UK presented by illegal content present on U2U services and by the use of such services for the commission or facilitation of priority offences. 

B:  This zone contains harm-related duties (identified in yellow): Ofcom Risk Profiles, several service provider risk assessment duties framed by reference to harm, plus the one substantive Section 10 duty framed by reference to harm (fed by the results of the harm-related risk assessment duties). Harm has its standard meaning in the Act: physical or psychological harm. 

C: This zone contains two service provider risk assessment duties which are independent of and unrelated to risk of harm, but which feed directly into a corresponding substantive Section 10 duty. 

D: This zone contains the substantive Section 10 duties: one based on harm and three which stand alone. Those three are not directly coupled to the service provider’s risk assessment.

This web of duties is undeniably complex. One can sympathise with the challenge of rendering it into a practical and readily understandable risk assessment process capable of feeding the substantive duties.  Nevertheless, a plainer and more consistently applied approach to terminology in Ofcom's documents would have paid dividends.

Safe speech by design

Proponents of a duty of care for online platforms have long dwelt on the theme of safety by design. It has come to the fore again recently with the government’s publication of a draft Statement of Strategic Priorities (SSP) for Ofcom under the Online Safety Act.  Safety by Design is named as one of five key areas.

Ofcom is required to have regard to the final version of the SSP in carrying out its functions under the Act. Given Ofcom’s regulatory independence the government can go only so far in suggesting how Ofcom should do its job. But, in effect, the SSP gives Ofcom a heavy hint about the various directions in which the government would like it to go.

So what does safety by design really mean? How might it fit in with platform (U2U) and search engine duties under the Online Safety Act (OSA)?

Before delving into this, it is worth emphasising that although formulations of online platform safety by design can range very widely [1] [2], for the purposes of the OSA safety by design has to be viewed through the lens of the specific duties imposed by the Act.

This piece focuses on the Act’s U2U illegality duties. Three of the substantive duties concern design or operation of the service:

• Preventing users encountering priority illegal content by means of the service (S. 10(2)(a))
• Mitigating and managing the risk of the service being used for the commission or facilitation of a priority offence (as identified in the most recent illegal content risk assessment of the service) (S. 10(2)(b))
• Mitigating and managing the risks of physical or psychological harm to individuals (again as identified in the most recent illegal content risk assessment) (S. 10(2)(c))

Two further substantive illegality duties are operational, relating to:

• Minimising the length of time for which priority illegal content is present on the service (S. 10(3)(a))
• Swiftly taking down illegal content where the service provider is alerted to or otherwise becomes aware of its presence. (S. 10(3)(b))

S.10(4) of the Act provides examples of the areas of design, operation and use of a service to which the duties apply and, if proportionate, require the service provider to take or use measures. Those include “design of functionalities, algorithms and other features.”

Safety by design in the Online Safety Act

When applied to online speech, the notion of safety by design prompts some immediate questions: What is safety? What is harm?

The OSA is less than helpful about this. It does not define safety, or safety by design. It defines harm as physical or psychological harm, but that term appears in only one of the five substantive illegality duties outlined above. Harm has more a pronounced, but not exclusive, place in the prior illegal content risk assessment that a platform is required to undertake.

Safety by design gained particular prominence with a last-minute House of Lords addition to the Bill: an introductory ‘purpose’ clause. This amendment was the result of cross-party collaboration between the then Conservative government and the Labour Opposition.

What is now Section 1 proclaims (among other things) that the Act provides for a new regulatory framework which has the:

“general purpose of making the use of internet services regulated by this Act safer for individuals in the United Kingdom.”

It goes on that to achieve that purpose, the Act (among other things):

“imposes duties which, in broad terms, require providers of services regulated by this Act to identify, mitigate and manage the risks of harm (including risks which particularly affect individuals with a certain characteristic) from

(i) illegal content and activity, and

(ii) content and activity that is harmful to children, …”

Finally, and most relevantly, it adds that:

“Duties imposed on providers by this Act seek to secure (among other things) that services regulated by this Act are … safe by design…”.

A purpose clause is intended to assist in the interpretation of the legislation by setting out the purposes for Parliament intended to legislate, rather than leaving the courts to infer them from the statutory language.

Whether such clauses in fact tend to help or hinder is a matter of lawyerly debate. This clause is especially confusing in its use of terms that are not defined by the Act and do not have a clear and obvious ordinary meaning (“safe” and “safe by design”), mixed up with terms that are specifically defined in the legislation (“harm”, meaning physical or psychological harm).

One thought might be that “safe” means safe from physical or psychological harm, and that “safe by design” should be understood accordingly. However, that seems unlikely since four of the five substantive illegality duties on service providers relate to illegal content and activity per se, irrespective of whether they might involve a risk of physical or psychological harm to individuals.

S.235 defines Ofcom’s “online safety functions” in terms of all its functions under the Act. In contrast, the transitional provisions for Video Service Providers define “safety duties” in terms focused on platform duties in respect of illegality and harm to children.

Similarly, in the earlier part of the Act only those two sets of duties are described (albeit merely in the section headings) as “safety duties”. “Safe by design” may possibly refer to those duties alone.   

The concept of safety by design tends to embody some or all of a number of elements: risk-creating features; prevention and reduction of harm; achieving those by appropriate design of a risk-creating feature, or by adding technical safeguards.

The most general aspect of safety by design concerns timing: that safety should be designed in from the outset rather than thought about afterwards.

Prevention itself has a temporal aspect, but that may relate as much to the kind of measure as to the stage of development at which it should be considered. Thus the Minister’s introduction to the Statement of Strategic Priorities says that it:

“includes ensuring safety is baked into platforms from the start so more harm is caught before it occurs”.

This could refer to the point at which a safety measure intervenes in the user’s activity, as opposed to (or as well as) the stage at which the designers consider it.

Later in the Statement, safety by design is expressly said to include deploying technology in content moderation processes. Providers would be expected to:

“…embed proportionate safety by design principles to mitigate the [risk of their service being used to facilitate illegal activity]. This should include steps such as … where proportionate, deploying technology to improve the scale and effectiveness of content moderation, considering factors including providers’ capacity and users’ freedom of expression and privacy rights.”

An analogy with product safety could suggest that safety by design is about identifying risk-creating features at the design stage and either designing those features in the safest way or incorporating safeguards. That aspect is emphasised by Professor Lorna Woods in a recent paper [3]:

“The objective of ‘safety by design’ is – like product safety – to reduce the tendency of a given feature or service to create or exacerbate such issues.”

Applied to products like cars that would mean that you should consider at the outset where safely to position the fuel tank, not unthinkingly place it somewhere dangerous and try to remedy the problem down the line, or after an accident has happened. Or, if a piece of machinery has a sharp cutting blade, consider at the outset how to add a guard into the design. A culture of safety by design should help to ensure that potential safety risks are considered and not overlooked.  

However, a focus on risk-creating features gives rise to particular difficulties when safety by design is translated to online platforms.

The underlying duty of care reasons for this have been rehearsed on previous occasions (here and here). In short, speech is not a tripping hazard, nor is it a piece of machinery. A cutting machine that presents a risk of physical injury to its operator is nothing like a space in which independent, sentient human beings can converse with each other and choose what to say and do.

Professor Woods [3] suggests that ‘by design’ seeks to ensure that products respect the law (my emphasis). If that is right, then by the same token it could be said that safety by design when applied to online platforms seeks to ensure that in their communications with each other users respect the law (or boundaries of harm set by the legislation). That is a materially different exercise, for which analogies with product safety can be taken only so far.

The June 2021 DCMS/DSIT paper Principles of safer online platform design opened with the statement that:

“Online harms can happen when features and functions on an online platform create a risk to users’ safety.”

For the illegality duties imposed by the OSA, when we set about identifying concrete features and functionalities that are said to create or increase risk of illegality, we run into problems when we move beyond positive platform conduct such as recommender and curation algorithms.

The example of recommender and curation algorithms has the merit of focusing on a feature that the provider has designed and which can causally affect which user content is provided to other users.

But the OSA duties of care – and thus safety by design - go well beyond algorithmic social media curation, extending to (for instance) platforms that do no more than enable users to post to a plain vanilla discussion forum.

Consider the OSA safety duties concerning priority illegal content and priority offences.  What kind of feature would create or increase a risk of, for example, an online user deciding to offer boat trips across the Channel to aspiring illegal immigrants?

The further we move away from positive content-related functionality, the more difficult it becomes to envisage how safety by design grounded in the notion of specific risk-creating features and functions might map on to real-world technical features of online platforms.

The draft SSP confirms that under the OSA safety by design is intended to be about more than algorithms:

“When we discuss safety by design, we mean that regulated providers should look at all areas of their services and business models, including algorithms and functionalities, when considering how to protect all users online. They should focus not only on managing risks but embedding safety outcomes throughout the design and development of new features and functionalities, and consider how to make existing features safer.”

Ofcom faced the question of risk-creating features when preparing the risk profiles that the Act requires it to provide for different kinds of in-scope service. For the U2U illegality risk profile it has to:

“carry out risk assessments to identify and assess the following risks of harm presented by [user to user] services of different kinds—

(a) the risks of harm to individuals in the United Kingdom presented by illegal content present on regulated user-to-user services and by the use of such services for the commission or facilitation of priority offences; …”

The risks that Ofcom has to identify and assess, it should be noted, are not the bare risk of illegal content or illegal activity, but the risk of harm (meaning physical or psychological harm) to individuals presented by such content or activity.

Ofcom is required to identify characteristics of different kinds of services that are relevant to those risks of harm, and to assess the impact of those kinds of characteristics on such risks. “Characteristics” of a service include its functionalities, user base, business model, governance and other systems and processes.

Although a platform has to carry out its own illegal content risk assessment, taking account of Ofcom’s risk profile, the illegality risks that the platform has to assess also include bare (non-harm-related) illegality.

Ofcom recognises that functionalities are not necessarily risk-creating:

“Functionalities in general are not inherently positive nor negative. They facilitate communication at scale and reduce frictions in user-to-user interactions, making it possible to disseminate both positive and harmful content. For example, users can engage with one another through direct messaging and livestreaming, develop relationships and reduce social isolation. In contrast, functionalities can also enable the sharing of illegal material such as livestreams of terrorist atrocities or messages sent with the intent of grooming children.” [6W.16]

Ofcom overcomes this issue in its proposed risk profiles by going beyond characteristics that of themselves create or increase risks of illegality. This is most clearly expressed in Volume 2 of its Illegal Harms Consultation:

“We recognise that not all characteristics are inherently harmful; we therefore use the term ‘risk factor’ to describe a characteristic for which there is evidence of a risk of harm to individuals. For example, a functionality like livestreaming is not inherently risky but evidence has shown that it can be abused by perpetrators; when considering specific offences such as terrorism or CSEA, a functionality like livestreaming can give rise to risk of harm or the commission or facilitation of an offence.” [5.26]

General purpose functionality and features of online communication can thus be designated as risk factors, on the basis that there is evidence that wrongdoers make use of them or, in some instances, certain combinations of features.

Since measures focused on general purpose features are likely to be vulnerable to objections of disproportionate interference with freedom of expression, for such features the focus of preventing or mitigating the identified risk is more likely to be on other aspects of the platform’s design, on user options and controls in relation to that feature (e.g. an option to disable the feature), or on measures such as content moderation.

Ofcom implicitly recognises this in the context of livestreaming:

“6.11 We acknowledge that some of the risk factors, which the evidence has demonstrated are linked to a particular kind of illegal harm, can also be beneficial to users. This can be in terms of the communication that they facilitate, or in some cases fulfilling other objectives, such as protecting user privacy. …

6.13 While livestreaming can be a risk factor for several kinds of illegal harm as it can allow the real-time sharing of illegal content, it also allows for real-time updates in news, providing crucial information to a wide-range of individuals.

6.14 These considerations are a key part of the analysis underpinning our Codes measures.”

The result is that while the illegality risk profiles that Ofcom has proposed include as risk factors a range of platform features that could be viewed as general purpose, they tend not to translate into recommended measures aimed at inhibiting that feature.

Here is a selection of features included in the proposed illegality risk profile:

Service feature	Risk (likelihood of increased risk of harm related to offences involving):
Ability to create user profiles	Grooming, harassment, stalking, threats, abuse, drugs and psychoactive substances, unlawful immigration, human trafficking, sexual exploitation of adults; and for the risk of fake profiles: Grooming, harassment, stalking, threats, abuse, controlling or coercive behaviour, proceeds of crime, fraud and financial services, foreign interference offences.
Users can form user groups	Grooming, encouraging or assisting suicide or serious self-harm, drugs and psychoactive substances, unlawful immigration, human trafficking.
Livestreaming	Terrorism, grooming, image-based CSAM, encouraging or assisting suicide or serious self-harm, harassment, stalking, threats, abuse.
Direct messaging	Grooming and CSAM, hate, harassment, stalking, threats, abuse, controlling or coercive behaviour, intimate image abuse, fraud and financial services offences.
Encrypted messaging	Terrorism, grooming, CSAM, drugs and psychoactive substances, sexual exploitation of adults, foreign interference, fraud and financial services offences.
Ability to comment on content	Terrorism, grooming, encouraging or assisting suicide or serious self-harm, hate, harassment, stalking, threats, abuse.
Ability to post images or videos	Terrorism, image-based CSAM, encouraging or assisting suicide or serious self-harm, controlling or coercive behaviour, drugs and psychoactive substances, extreme pornography, intimate image abuse.
Ability to repost or forward content	Encouraging or assisting suicide or serious self-harm, harassment, stalking, threats, abuse, intimate image abuse, foreign interference.
Ability to search for user generated content	Drugs and psychoactive substances, firearms and other weapons, extreme pornography, fraud and financial services offences.
Hyperlinks	Terrorism, CSAM URLs, foreign interference offences.

Other functionality risk factors include anonymity, user connections (such as friending and following), group messaging, and ability to post or send location information.

Designation of general purpose functionality as a risk factor reaches a high point with hyperlinks. Since terrorists and other potential perpetrators can use hyperlinks to point people to illegal material, hyperlinks can be designated as a risk factor despite not being inherently harmful.

It is worth recalling what the ECtHR said in Magyar Jeti Zrt (ECtHR) about the central role of hyperlinks in internet communication:

“Furthermore, bearing in mind the role of the Internet in enhancing the public’s access to news and information, the Court points out that the very purpose of hyperlinks is, by directing to other pages and web resources, to allow Internet users to navigate to and from material in a network characterised by the availability of an immense amount of information. Hyperlinks contribute to the smooth operation of the Internet by making information accessible through linking it to each other.”

General purpose functionality as a risk factor was foreshadowed in the June 2021 DCMS paper. Arguably it went further, asserting in effect that providing a platform for users to communicate with each other is itself a risk-creating activity:

          “Your users may be at increased risk of online harms if your platform allows them to:

• interact with each other, such as through chat, comments, liking or tagging
• create and share text, images, audio or video (user-generated content)”

In the context of the internet in the 21st century, this list of features describes commonplace aspects of the ability to communicate electronically. In a former age we might equally have said that pen, paper, typewriter and the printing press are risk factors, since perpetrators of wrongdoing may use written communications for their nefarious purposes.

Whilst Ofcom recognises the potential freedom of expression implications of treating general purpose functionalities as illegality risk factors, it always has to be borne in mind that from a fundamental rights perspective the starting point is that speech is a right, not a risk. Indeed the Indian Supreme Court has held that the right of freedom of expression includes the reach of online individual speech:

"There is no dispute that freedom of speech and expression includes the right to disseminate information to as wide a section of the population as is possible."

That is not to suggest that freedom of expression is an absolute right. But any interference has to constitute a sufficiently clear and precise rule (especially from the perspective of the user whose expression is liable to be interfered with), then satisfy necessity and proportionality tests.

Preventative technological measures

A preventative approach to safety by design can easily lean towards technological measures: since this is a technology product, technological preventative measures should be designed in to the service and considered at the outset.

Professor Woods [3], argues that:

“Designing for safety (or some other societal value) does not equate to techno-solutionism (or techno-optimism); the reliance on a “magic box” to solve society’s woes or provide a quick fix.”

However, in the hands of government and regulators it has a strong tendency to do so.[4].  Indeed the draft SSP devotes one of its five key priorities to Technology and Innovation, opening with:

“Technology is vital to protecting users online and for platforms fulfilling their duties under the Act.”

Later:

“It is not enough that new, innovative solutions to known problems exist – online service providers must also adopt and deploy these solutions to improve user safety. … The government … encourages Ofcom to be ambitious in its [code of practice] recommendations and ensure they maintain pace with technology as it develops.”

We have already seen that in the draft SSP, safety by design is said to include deploying technology in content moderation processes.

On the basis of prevention, an inbuilt technological design measure that reduces the amount of (or exposure to) illegal user speech or activity should be preferable to hiring legions of content moderators when the platform starts operating.

However, translating duties of care or safety by design into automated or technology-assisted content moderation can come into conflict with an approach in which non-content-specific safety features are seen as preferable.

Professor Woods said in the same paper:

“At the moment, content moderation seems to be in tension with the design features that are influencing the creation of content in the first place, making moderation a harder job. So, a “by design” approach is a necessary precondition for ensuring that other ex post responses have a chance of success.

While a “by design” approach is important, it is not sufficient on its own; there will be a need to keep reviewing design choices and updating them, as well as perhaps considering ex post measures to deal with residual issues that cannot be designed out, even if the incidence of such issues has been reduced.”

As to what ex post measures might consist of, in a letter to The Times in August, Professor Woods said:

“Through a duty of care, service operators are required to ensure that their products are as safe as reasonably possible and to take steps to mitigate unintended consequences. Essentially this is product safety, or health and safety at work. This approach allows a range of interventions that do not rely on content take-down and, indeed, could be content-neutral. One example might be creator reward programmes that incentivise the spreading of clickbait material. (emphasis added)].

Maeve Walsh, writing for the Online Safety Network shortly before publication of the draft SSP [5], contrasted safety by design with thinking about the OSA “primarily as a takedown-focused regime, centering on individual pieces of content.”

Content-neutrality suggests that a safety measure in relation to a functional feature should, rather than relating specifically to some kind of illegal or harmful content, either have no effect on content as such or, if it does affect user content, do so agnostically.

Some measures have no direct effect on user content: a help button would be an example. Others may affect content, but are not targeted at particular kinds of content: for instance, a friction-reducing measure like capping the permissible number of reposts, or other measures inhibiting virality.

A measure such as a quantititive cap on the use of some feature has the advantage from a rule of law perspective that it can be clearly and precisely articulated. However, by virtue of the fact that it constrains legitimate as well as illegitimate user speech across the board, it is potentially vulnerable to proportionality objections.

Thanks to the difficulty of making accurate illegality judgements, automated content filtering and blocking technologies are potentially at risk on both scores.

[1] Trust & Safety Professional Association. Safety by Design Curriculum chapter.

[2] Australian eSafety Commissioner. Safety by Design.

[3] Professor Lorna Woods, for the Online Safety Network (October 2024). Safety by Design. 

[4] Maria P. Angel, danah boyd (12 March 2024). Proceedings of 3rd ACM Computer Science and Law Symposium (CSLAW’24) Techno-legal Solutionism: Regulating Children’s Online Safety in the United States.

[5] Maeve Walsh, for the Online Safety Network (11 October 2024). Safety by design: has its time finally come?