Source · Select Committees · Science, Innovation and Technology Committee
Recommendation 29
29
Accepted
Mandate generative AI platforms to automatically label AI-generated media with metadata and watermarks.
Recommendation
To effectively tackle amplified misinformation as per Principle 1, the government should work with relevant experts and platforms to develop technology that automatically detects AI-generated media, meeting mis/ disinformation at its source. It should mandate all generative AI platforms, and platforms that employ generative AI technologies, to automatically label AI-generated media with metadata and visible watermarks that cannot be removed. (Recommendation, Paragraph 78) Digital advertising market
Government response summary AI-generated
The government responds by highlighting its past initiative, the Deepfake Detection Challenge, which brought together experts to develop solutions for detecting synthetic media. However, it does not commit to working with platforms to develop new detection technology or mandating automatic labelling of AI-generated media with metadata and watermarks.
Summary of the government's response below — read the verbatim text to verify.
Government Response
Accepted
HM Government · verbatim extract
Accepted
As previously mentioned, government already engages with relevant experts to understand the potential for identifying AI-generated content, particularly where it constitutes harmful content. We are working with partners to investigate the potential for detecting AI-generated content to support transparency of synthetically generated content, this includes assessing both evidence on the feasibility of technical solutions and levers to ensure that these technologies are developed and deployed in a beneficial way. Together with Home Office, we ran the Deepfake Detection Challenge to understand how government, academia and industry can work together in the development of innovative and practical solutions to detect synthetic media. The Challenge examined existing capabilities and identified gaps requiring innovative approaches and novel solutions. Follow-up work from the challenge is underway. We continue to explore solutions that enable users and institutions to determine what media is real and what is AI-generated, this is a key part of tackling a wide range of AI risks. Ofcom are also undertaking research to assess the merits and limitations of different measures to identify deepfakes, including watermarking tools, provenance metadata schemes, AI labels, and context annotations. While the findings are not new rules or expectations for tech firms, they can be used to guide providers or those deploying these tools to help identify deepfake content. Ofcom’s research will also inform their future policy development and supervision of regulated services under the Online Safety Act.
Read the full response on Parliament ↗