Ofcom hash matching for AI deepfake intimate images: 30 September 2026

Ofcom's statement of 10 September 2026 says online platforms should use hash matching to detect and prevent the sharing of non-consensual intimate images, including AI-generated explicit deepfakes, by 30 September 2026, or show equally effective alternatives. I could read it only as republished by Wired-Gov, because Ofcom's own page returned an error, so check Ofcom directly.

What the statement says

It describes an enforcement programme with SWGfL and StopNCII.org, fines of up to 10% of global revenue, a consultation by the end of the year on a 48-hour removal requirement, and a progress report planned for summer 2027.

Who this is aimed at

The statement addresses online platforms that host content, not people who make a promotional video. A brand video team is unlikely to be the target, but the context matters: platforms you publish to are tightening enforcement on deepfakes of real people.

What a video team can do

Use only faces and voices you have rights to. Sume Face Swap is a Beta endpoint that takes an avatar_handle for a ready avatar plus a source video, so the identity is one you set up as an avatar rather than any person's photo. Keep the consent record next to the job id.

This is a summary of a public statement, not legal advice.