Overview
- The Home Office will partner with Microsoft, academics and independent experts to design a shared evaluation framework for spotting synthetic media.
- The programme will stress-test detection technologies against threats such as sexual abuse, fraud and impersonation to identify where tools fall short.
- Government figures estimate about eight million deepfakes were shared in 2025, up from roughly 500,000 in 2023.
- Police leaders say consistent benchmarks will help protect victims and improve investigators’ ability to keep pace with offenders.
- The announcement follows ongoing probes into X and its Grok chatbot over sexualised deepfakes and comes after the UK criminalised the creation of non-consensual intimate images.