OpenAI and Anthropic push for binding AI regulations in UK

Representatives from OpenAI and Anthropic testified before a British parliamentary committee, urging the introduction of binding regulations for AI safety, a stance that contrasts with Google DeepMind and Meta, which opposed mandatory legislation. The Business, Innovation, Science, and Trade Committee convened a hearing with executives, former employees, and civil servants, highlighting divergent views on AI governance. Anthropic’s Liam Booth-Smith emphasized the need for statutory rules, stating that voluntary commitments are insufficient. OpenAI’s Tom Duff Gordon supported a statutory framework and international convention, advocating for mandatory incident reporting. In contrast, Google DeepMind’s Owen Larter and Meta’s Rob Sherman argued for reliance on existing regulatory bodies. Meanwhile, MI5 director Ken McCallum warned of AI’s potential to enhance bioweapon capabilities, underscoring the urgency for stricter oversight. The UK’s AI Security Institute (AISI) relies on company cooperation for model testing, with limited enforcement power. Anthropic’s Booth-Smith noted restricted access to its models for UK AISI, citing U.S. government involvement. Committee chair Liam Byrne called for a global AI convention during the UK’s G20 presidency, emphasizing the need for a safety framework that transcends current political boundaries.