Tom Duff Gordon, OpenAI's head of policy in EMEA, calls on the UK to introduce binding AI rules to "establish durable, mandatory, capability-based requirements"

Binding AI safety rules could soon be mandated for the most advanced AI models developed in the UK.

OpenAI's head of policy for Europe, the Middle East, and Africa, Tom Duff Gordon, is advocating for the U.K. to implement binding regulations for advanced AI development. He argues that the U.K. government should seize the current political opportunity to establish mandatory, capability-based safety requirements for "frontier AI" models. Duff Gordon suggests that such legislation should include compulsory independent third-party testing, a strong emphasis on cybersecurity, and protocols for monitoring and reporting safety incidents. He believes this would align the U.K. with regulatory approaches in the EU and the U.S., positioning it as a leader in AI governance. OpenAI's proposal comes amid increasing pressure on the Labour government to address AI risks, particularly after a researcher's public accusations against major AI companies regarding safety concerns. The company has previously expressed support for mandatory, capability-based national AI safety regulation.

OpenAI's endorsement of mandatory U.K. AI legislation suggests a strategic move to shape the regulatory landscape for frontier AI, potentially setting a precedent for other nations. By calling for capability-based requirements and independent testing, the company appears to be preempting stricter, potentially more burdensome, regulations that could arise from a less informed, reactive approach. This advocacy also positions OpenAI as a responsible actor, keen to collaborate with governments while guiding the development of AI safety standards, thereby influencing the pace and direction of AI innovation.

The U.K. government's AI Safety Institute is highlighted as a potential central body for enforcing these new regulations, indicating a preference for leveraging existing expertise. This approach, coupled with a call for international interoperability, signals a desire to harmonize global AI governance frameworks, making it easier for companies like OpenAI to operate across different jurisdictions. It also suggests a potential shift from the current reliance on voluntary commitments to a more structured, government-led oversight of AI development and deployment.