Microsoft President Backs UK Leadership in Global AI Regulation

Smith described AI as both a weapon and a tool capable of undermining global cybersecurity. He identified Britain’s AI Security Institute, launched in 2023, as a uniquely positioned leader in safety testing. Unlike the European Union, which often faces skepticism in Washington, the U.K. can operate as a neutral arbiter. Despite reports that the Trump administration requested major labs like OpenAI and Anthropic withhold models from British testers, the institute confirmed its continued access to these frontier technologies.

While major tech leaders recently signed a "morally binding" safety agreement at the White House, Smith dismissed voluntary self-regulation as insufficient. He warned that such measures only bind responsible actors, leaving the door open for less scrupulous firms to bypass safety protocols. Smith proposed concrete mandates, including legal obligations to report security breaches by rogue AI agents and the implementation of mandatory "kill-switches" to halt systems that escape human control. By advocating for these external evaluations, he hopes to shift the industry toward a framework where independent auditors hold the final say before any new product reaches the public.