TL;DR
Ofcom set out tougher Online Safety Act duties on child sexual abuse material (CSAM) on 16 November, saying AI has made the problem “significantly worse”. Platforms will have to ban offending users, run hash matching on more services and, on file-sharing services and the biggest social networks, use automated tools to find abuse images that have never been seen before. The duties are expected to apply from early 2027, subject to Parliament.
The regulator already requires services to assess their CSAM risk and take proportionate steps. The new package, which follows a consultation, goes further on three fronts.
The three new duties
First, any user who creates, uploads or shares abuse content must be removed from the service. Second, hash matching, which compares files against a database of known abuse images, extends to more sites and apps, pornography services among them. Third, every file-sharing service and the largest social networks must use automated tools that proactively detect and remove CSAM not yet on any list.
Ofcom also tightened grooming protections. Services covered by its children’s codes must either run robust age verification to identify children’s accounts, or extend their anti-grooming safeguards to all users. Some platforms must take more care before their algorithms recommend content to children.
Why AI features in the decision
In its notes, Ofcom cites Internet Watch Foundation research from the IWF’s AI CSAM Report 2026, which recorded a 26,385% annual rise in AI-made abuse videos online. The NSPCC put the UK figure at roughly 100 abuse-image crimes recorded by police every day, based on its own reading of force data.
“We are not satisfied that enough is being done,” said Oliver Griffiths, who leads Ofcom’s online safety work, warning that firms which fail to act “should expect to face serious consequences”.
The figures add to the Childlight data we covered in our report on AI-related abuse material.
Looking forward
Ofcom dropped planned rules on children livestreaming, because the under-16 social media ban due next year will require many services to age-check users out altogether. It is also moving faster elsewhere: relevant services are being told to hash-match known terrorist content too, and Ofcom plans to consult by spring on directing platforms to adopt specific detection technologies rather than assessing the options themselves. In our view, the move from platforms assessing tools for themselves to Ofcom naming the tools is the change operators should plan around.