OpenAI and Anthropic Urge UK Parliament to Enact Binding AI Safety Laws

OpenAI and Anthropic told a UK parliamentary committee this week that voluntary safety pledges are insufficient for frontier AI systems and called for binding legislation. Anthropic's Europe head Liam Booth-Smith stated the company has "consistently believed that there should be binding rules on companies," while OpenAI's Tom Duff Gordon backed mandatory incident reporting and an international convention.

The hearing before the Business, Innovation, Science, and Trade Committee exposed a sharp industry divide. Google DeepMind's Owen Larter and Meta's Rob Sherman declined to support new statutes, pointing instead to existing regulators and technical standards. Committee chair Liam Byrne said the session aimed to "light up the wide differences in opinions between some of the wealthiest firms on Earth."

Security chiefs sound alarm on bioweapon risk

Hours later, MI5 director Ken McCallum warned that AI's greatest misuse potential lies in "changing the stakes on existing or even new bioweapons." The juxtaposition underscored a regulatory gap: Britain's AI Security Institute only tests models when companies voluntarily submit them, leaving critical capabilities unexamined.

Why it matters for GPU / AI infrastructure: Binding safety rules will likely mandate compute governance, audit trails, and controlled access to high-capacity training clusters. Infrastructure providers must prepare for compliance workloads, data residency requirements, and certified deployment environments that align with emerging statutory frameworks.