When does protecting someone from harm become a legal duty to stop them from seeing something? That ethical and legal tension lies at the centre of a new amendment to the UK’s Online Safety Act, which now requires tech companies to prevent certain self-harm content from appearing on their platforms rather than merely reacting after users flag it. The change tightens one of the most contested pieces of digital regulation in recent British history, promising stronger safeguards for vulnerable users while reopening familiar debates about free expression, privacy, and technical feasibility.
How the Online Safety Act amendment changes platform duties
Passed as part of a broader push to make the internet safer, the Online Safety Act originally obliged platforms to remove illegal material and take proportionate steps to limit harmful but lawful content. The amendment goes further: content that encourages, instructs, or depicts self-harm in particular ways is reclassified so platforms must put preventive systems in place to stop it appearing in the first place. Ofcom, the UK communications regulator, will enforce the new duties and retain powers to issue fines and require compliance plans.
Government and charity briefings say the aim is clear — shield young and vulnerable users from graphic or instructional self-harm material that can normalise or enable suicidal behaviour. Organisations such as Samaritans and the NSPCC have welcomed the tougher stance, citing decades of research linking certain online portrayals of self-harm to copycat behaviour and heightened risk among adolescents. Their argument: prevention — keeping dangerous material from gaining visibility — must complement clinical and educational interventions.
The practical challenges of preventing self-harm content
Despite its intentions, the amendment revives long-standing practical and ethical questions. Automated classifiers — the primary tools platforms use to identify content at scale — struggle with nuance. Machine learning models can flag explicit imagery or commonly associated phrases, but context is crucial. A news report about suicide prevention and a how‑to video encouraging self‑harm may use overlapping vocabulary yet serve opposite social functions. False positives risk erasing helpful resources; false negatives mean the law’s objectives go unmet.
The amendment also collides with technical realities around encryption and private messaging. Platforms offering end‑to‑end encrypted services face a dilemma: how to prevent self-harm content without breaking encryption or user privacy. Some companies could consider client‑side scanning or other invasive techniques to comply, raising serious privacy concerns among rights advocates. For smaller firms and niche communities, the costs of building sophisticated moderation systems may be prohibitive, potentially forcing them out of the UK market or into expensive third‑party services.
Rights, definitions and the risk of overreach
Civil liberties groups and technologists warn that a duty to “prevent” content may be interpreted expansively, sweeping up legitimate expression: journalism, first‑person recovery narratives, academic analysis, or peer support groups that discuss self‑harm to help rather than harm. Legal experts note that Ofcom already has wide discretion under the Online Safety Act to define “harmful” content and proportionate measures. How regulators choose to interpret the amendment — and how courts balance the protection of life against freedom of expression under the European Convention on Human Rights — will shape enforcement outcomes.
There is a very real danger of unintended consequences. Over‑removal could drive vulnerable conversations into darker corners of the internet — encrypted apps, closed forums, or anonymised networks — where users receive less support and interventions are harder. Under‑enforcement, conversely, would fail the people the law seeks to protect.
What platforms and policymakers should consider
Platform responses have been mixed. Major companies have urged clear, objective rules and predictable timelines for compliance; vague or shifting obligations complicate implementation. Digital rights organisations are calling for narrow, evidence‑based definitions, robust oversight, and transparent safeguards to prevent mission creep and preserve access to legitimate support and research resources.
Concrete steps that could reduce harm without silencing critical conversation include:
– Clear carve‑outs for journalistic, academic, and therapeutic material.
– Independent audits of moderation decisions and published error rates for automated detection.
– Rules on the acceptable technical approaches to monitoring encrypted content to protect privacy.
– Funding and support for smaller platforms to meet compliance without exiting the market.
– Parallel investment in offline mental‑health services, school programmes, and crisis hotlines.
Policymakers argue that a legal requirement to prevent the most dangerous material will force firms to design safety into product architecture — recommendation engines, search, and algorithms — instead of treating safety as an afterthought. From a platform perspective, however, the operational and financial burdens are substantial. Even large companies that invest heavily in moderation teams and automated tools may still face regulatory risk and the social cost of over‑cautious filtering.
Conclusion: balancing protection and expression under the Online Safety Act
The amendment to the Online Safety Act signals a broader international trend: governments increasingly expect tech companies to embed prevention into systems rather than simply remove content after the fact. Whether this shift will save lives, stifle speech, or do both depends on definitions, technologies, and the safeguards regulators and platforms adopt. The central question remains whether it is possible to design rules and systems that protect without silencing — that prevent harm without erasing hope. Robust transparency, narrowly tailored rules, and investments in both online and offline support must accompany enforcement if the Online Safety Act is to achieve its aim without unintended harms.
