UK regulator backs automated systems to detect explicit deepfakes  

The UK is moving toward more proactive detection of AI-generated intimate abuse, with media regulator Ofcom urging platforms to deploy automated hash-matching systems to identify and block explicit deepfakes shared without consent.

Disproportionately affecting women and girls, the spread of such imagery surged with AI-generated deepfakes. Officials in France, Germany and Ireland, among others, criticized Elon Musk’s X for its alleged sexualized deepfakes and child sexual abuse material.

“Given the urgent need to better protect women and girls online, we are now adding a recommendation to our codes that certain sites and apps expand their use of automated technology – known as ‘hash matching’ – to detect illegal intimate images shared without consent, such as explicit deepfakes,” Ofcom said in a release.

Hash matching is a technology that converts harmful images into digital fingerprints known as “hashes.” These hashes are stored in a database to match against further attempts to upload the same or similar versions of the image.

Hash-matching systems are effective at identifying previously flagged material, though regulators and platforms continue to face challenges as generative AI tools produce increasingly varied synthetic content at scale. Unlike traditional image abuse, generative AI systems can rapidly create modified or entirely synthetic variants designed to evade existing detection databases.

The move reflects a broader regulatory shift toward automated detection infrastructure as governments struggle to contain the rapid spread of synthetic intimate imagery and AI-powered impersonation tools.

Ofcom named StopNCII as a recommended hash database service. The UK regulator expects the intimate image abuse amendments to its Illegal Content Codes to take effect this autumn, pending parliamentary process. These codes apply to user-to-user services as well as to search.

The new code recommending the use of automated detection technology and hashes comes ahead of new legislation which will introduce a ban on nudification tools and requirements for non-consensual intimate images to be taken down within 48 hours.

In guidance for additional safety measures for online safety and detecting intimate image abuse, Ofcom is recommending “proactive technology” such as hash matching. Such technologies can be used as part of a content moderation system, the watchdog advises.

Ofcom said platforms should combine hash matching, keyword detection, behavioural analysis and other automated moderation tools to identify and disrupt intimate image abuse while complying with legal limits around monitoring user-generated content.

As AI-generated abuse becomes easier to produce and distribute, regulators are increasingly treating automated detection systems and digital fingerprinting infrastructure as essential components of online safety enforcement.

Article Topics

AI fraud  |  deepfake detection  |  deepfakes  |  hash-matching  |  Ofcom  |  regulation