AI safety institute launches £8.5 million initiative to enhance systemic safety research

Up to £200,000 will be awarded to projects ensuring AI safety in critical sectors.
AI researchers are shifting focus from scaling up models to smarter, human-like problem-solving methods.

The AI Safety Institute is launching an £8.5 million funding scheme to support research on AI system safety, while the initiative will back studies on preventing unexpected failures in AI technologies and addressing challenges linked to their rapid deployment.

The Systemic Safety Grants Programme, run in partnership with the Engineering and Physical Sciences Research Council and Innovate UK, will initially fund around 20 projects. Each project can receive up to £200,000 to explore risks AI might present to society in the near future. Additional funding will follow as further phases are introduced.

Systemic AI safety focuses on the broader infrastructure supporting AI across sectors, including healthcare and energy. Ian Hogarth, chair of the institute, emphasised the importance of addressing risks in critical industries. He highlighted that diverse research teams will contribute to building essential knowledge about AI-related threats, such as deepfakes and system malfunctions.

Applications are open until 26 November, with successful projects to be announced by January 2025. Grants will be awarded the following month, supporting efforts to ensure AI systems remain safe, reliable, and trustworthy as their use expands across the economy.