UK to tackle deepfake fraud adverts this summer
Key Points
- Ofcom will publish a draft Fraudulent Advertising Code of Practice this summer, informed by its deepfake research.
- The regulator said 22% of internet users now report encountering fake or deceptive images or videos online.
- Ofcom research set out a dozen deepfake mitigation methods available today, from output filters to downranking host sites.
- An internal Ofcom experiment showed watermarks could be removed from images.
- The code forms part of a wider online safety policy consultation.
Ofcom will publish a draft Fraudulent Advertising Code of Practice this summer that draws on its research into deepfakes, as more than one in five internet users report encountering fake or deceptive images or videos online.
The regulator confirmed the timing in its strategic approach to AI for 2026/27, published on 4 June 2026. Ofcom said over 1 in 5 internet users, at 22%, now report encountering fake or deceptive images or videos online.
The regulator said its deepfake research has directly informed the development of the draft code, which it will publish as part of a wider online safety policy consultation in the summer.
Ofcom said it has been investigating the causes and consequences of a rise in harmful deepfakes online. The regulator first published a discussion paper, Deepfake Defences, that examined the nature of deepfakes and set out a typology covering those intended to demean, defraud and disinform.
Ofcom said the paper set out a range of mitigation options that could be deployed at different points in the AI lifecycle. These ranged from AI developers applying output filters to their models, to search engine providers taking action to downrank websites hosting deepfakes.
The regulator said the Deepfake Defences paper lists a dozen distinct mitigation methods that are available to deploy today.
Testing watermarks
Ofcom said its second discussion paper, The Attribution Toolkit, examined so called attribution measures, including AI labels and metadata schemes. The regulator said the paper included findings from an internal technical experiment on common watermarking schemes.
Ofcom said the experiment revealed the extent to which watermarks could be removed from images. The regulator said the paper identifies where some mitigation methods could be further improved, such as making watermarks more resistant to tampering.
Ofcom said it hopes AI developers and safety technology firms bear its findings in mind as they continue to iterate their deepfake safeguards.
How the new code fits in
Ofcom said it has been able to assess the merits of different techniques for preventing the creation of deepfake fraudulent adverts, drawing on its research. The regulator said this work has directly informed the draft Fraudulent Advertising Code of Practice.
Ofcom said its discussion papers also serve as a source of intelligence for others in the wider online safety community as they seek to tackle deepfakes. The regulator said AI generated content can cause harm even when it is not used maliciously, because it can create unintentional inaccuracies that reduce trust in content, especially news.
Ofcom said it will continue to act within its powers to address harms exacerbated by AI, including deepfake intimate image abuse and synthetic child sexual abuse material, and will take enforcement action where it identifies non compliance with the Online Safety Act.
