AI Firms Split Over Binding Safety Rules at UK Parliamentary Hearing
OpenAI and Anthropic backed binding rules; Google DeepMind and Meta did not endorse new legislation. UK officials said the AI Security Institute relies on voluntary model access and cannot block releases it considers unsafe. Anthropic said U.S. restrictions initially limited UK access to its Mythos 5.1 model, but access has since been granted. MI5 Director Ken McCallum warned AI could increase bioweapons risks.
OpenAI and Anthropic supported binding AI safety regulation at a UK parliamentary hearing. Google DeepMind and Meta did not endorse new legislation, highlighting differing views on governance of advanced AI. Anthropic said voluntary safety pledges are inadequate. OpenAI backed mandatory incident reporting and international rules; Google favored standards, while Meta cited existing regulators.
UK officials said the AI Security Institute depends on voluntary access to models and cannot stop companies from releasing models it considers unsafe. S. 1 model, but access has since been granted. MI5 Director Ken McCallum warned that AI could raise bioweapons risks.
Former AI researchers also questioned whether the industry can regulate itself.