OpenAI and Anthropic Tell British Lawmakers: Regulate Us

The Business, Innovation, Science, and Trade Committee, a cross-party panel of British MPs, held an evidence session on Tuesday, featuring representatives from four frontier AI companies, four former staff who have spoken out against their employers, and senior civil servants. Anthropic's head of Europe, Liam Booth-Smith, advocated for a statutory framework, stating that voluntary commitments are insufficient.

OpenAI's vice president, Tom Duff Gordon, also supported a statutory framework and an international convention, emphasizing the need for mandatory incident reporting. Google DeepMind's Owen Larter and Meta's Rob Sherman were less clear in their support for statutory regulation. The session highlighted a significant divide between AI leaders and British policymakers, underscoring the limitations of voluntary self-regulation.

Meanwhile, MI5 director Ken McCallum warned of the potential misuse of AI in bioweapon development during his annual threat update. The committee chair, Liam Byrne, stated that the committee aimed to expose the stark differences in opinions among the world's wealthiest firms and demonstrate the inadequacy of voluntary self-regulation.

The private sector's reluctance to provide full access to their models for assessment by the UK's AI Security Institute (AISI) further exemplified the tension between the parties. Former OpenAI employee Daniel Kokotajlo and Steven Adler, who left OpenAI in late 2024, expressed doubts about the companies' commitment to self-regulation.

Geoffrey Irving, also a former lab researcher, argued that the labs' belief that they are locked in a race uninterupted by governments is false, and that they have the option to cease. In conclusion, the hearing revealed the urgent need for a comprehensive and international approach to AI regulation, with Britain's G20 presidency as a potential opportunity for a collaborative global convention.