Quick answer
What you need to know
- Who removes it?
- The website does—not Ofcom. Software may help the site find or judge posts.
- Why mistakes happen
- Software can struggle with context, evidence, irony and language used by particular communities.
- Can I appeal?
- The useful version is prompt, easy to find and reviewed by someone who can understand the context.
Checked against official sources on 8 August 2026 sourcesource
Want the detail?Check the law, evidence and open questionsIncludes what the law requires, what Ofcom says, safeguards, technical diagrams and every source.
See the technical process diagram
What does the law actually require?
In the lawThe Act requires regulated services to assess illegal-content risks and use proportionate systems and processes. It also contains duties concerning freedom of expression and privacy; additional duties for the largest Category 1 services cover terms of service, user empowerment, journalistic content and content of democratic importance. source
Is this happening now?
What Ofcom saysIllegal-harms duties have been in force since 2025. In July 2026 Ofcom opened consultations on additional Category 1 duties, including complaints, user empowerment, journalistic content, freedom of expression and privacy assessments. source
What do we know so far?
What we foundThe regulatory model is systems-focused: risk assessment, mitigation, record-keeping and complaints processes. It does not make Ofcom the editor of each individual post. Automated classification may be one tool within those systems, but the right response depends on risk and service design. source
What could go wrong?
Our viewAt scale, providers may rely on automated moderation that cannot reliably understand context, irony, documentary evidence or reclaimed language. Mistakes can fall hardest on journalists, minorities, survivors and people discussing self-harm or abuse. Strong notice, explanation and appeal processes should be treated as safety features.
What is this trying to fix?
In the lawThe Act’s central rationale is to reduce exposure to illegal content and protect children from defined harmful content. Risk-based systems can address design features that amplify harm, rather than relying only on individual takedowns. source
What protections are built in?
In the lawProviders must have regard to users’ freedom of expression and privacy when deciding and implementing safety measures. Complaint and redress obligations vary by service category, with further detailed guidance still being developed for categorised services. source source
What has not been decided?
- Will services publish useful error rates and explain material automated decisions?
- Are appeals prompt, accessible and reviewed by people with relevant context?
- Can researchers inspect differential impacts without exposing user data?
- How will safety duties interact with private terms that go beyond the law?
Sources for this page
- Regulator statement
Ofcom. Ofcom's approach to implementing the Online Safety Act.
Ofcom's implementation roadmap, milestones, active duties and planned regulatory work, updated in July 2026. Published 26 October 2023; last checked 8 August 2026.
- Regulatory guidance
Ofcom. Online Safety Act compliance guide for providers of online services.
Practical regulatory guidance for services on deciding whether the Act applies and meeting their duties. Published 16 December 2024; last checked 8 August 2026.
- Primary legislation
UK Parliament. Online Safety Act 2023.
The current consolidated text of the Act, including regulated services, duties, enforcement powers, safeguards and exemptions. Published 26 October 2023; last checked 8 August 2026.