Question to the Home Office:

To ask the Secretary of State for the Home Department, how many potential crimes involving AI deepfake programmes were reported in each of the last three years.

Generative artificial intelligence services have made it easier to produce convincing deepfake content and, whilst there are legitimate use cases this is also impacting a range of crime types.

The Home Office is working closely with law enforcement, international partners, industry and across Government to address the risks associated with deepfakes. This includes reviewing the extent to which existing criminal law provides coverage of AI-enabled offending and harmful behaviour, including the production and distribution of deepfake material using generative AI. If the review suggests alterations to the criminal law are required to clarify its application to AI-generated synthetic and manipulated material then amendments will be considered in the usual way.

The Online Safety Act places new requirements on social media platforms to swiftly remove illegal content - including artificial intelligence-generated deepfakes - as soon as they become aware of it. The Act also updates Ofcom’s statutory media literacy duty to require it to take tangible steps to prioritise the public's awareness of and resilience to misinformation and disinformation online. This includes enabling users to establish the reliability, accuracy, and authenticity of content.

We have no current plans to ban services which generate deepfakes, however Government has been clear that companies providing AI services should take steps to ensure safety and reduce the risks of misuse. This was discussed at the Government’s AI Safety Summit in November 2023, reinforcing our commitment to international collaboration on this shared challenge.

Crime is recorded on the basis of the underlying offence, not whether a deepfake was involved, and we are therefore unable to provide a figure for deepfake-enabled crimes.

We are unable to provide figures for departmental spending as this is captured according to crime type, or broader work on artificial intelligence, and not broken down into activities specific to deepfakes.