OpenAI and Anthropic Tell British Lawmakers: Regulate Us
wired ·
OpenAI and Anthropic representatives testified before a British parliamentary committee that voluntary safety pledges are insufficient, advocating instead for binding statutory frameworks and international conventions. This stance created a clear divide with rivals Google DeepMind and Meta, which opposed new legislation in favor of existing standards or regulators. The hearing highlighted the limitations of the current UK regulatory regime, as civil servants admitted that the AI Security Institute relies entirely on company cooperation and lacks enforcement powers to prevent dangerous model releases. Simultaneously, MI5 director Ken McCallum warned of severe threats, including the potential for AI to facilitate the creation of new bioweapons. Whistleblowers from the industry reinforced these concerns, arguing that companies cannot be trusted to self-regulate due to competitive pressures. Committee chair Liam Byrne emphasized the need to expose these conflicting views among wealthy firms to justify stronger government intervention. Looking ahead, the committee plans to leverage Britain’s upcoming G20 presidency to broker an international AI safety convention independent of US political shifts.
wired