On 23 July 2024, the Office of Communications (Ofcom) announced the issuance of guidance on employing red teaming as a method to evaluate and mitigate harms posed by Generative AI (GenAI) technologies. The initiative is part of Ofcom's broader strategy to ensure the safety of online users in accordance with the Online Safety Act 2023, which mandates risk assessments and the implementation of proportionate steps to mitigate risks associated with illegal content or content harmful to children, including those arising from GenAI functionalities. The guidance outlines the process of red teaming, which involves simulating attacks on GenAI models to identify vulnerabilities that could lead to the generation of harmful content, and suggests measures to address these vulnerabilities. This approach is supported by major model developers and the AI Safety Institute, aligning with global efforts towards the safe deployment of AI technologies.
Original source