Microsoft President Brad Smith has urged nations to accelerate the creation of international AI regulatory frameworks. He noted that current oversight for AI companies is less stringent than that applied to consumer goods like milk or washing machines. Smith emphasized that AI poses new cybersecurity risks, necessitating clear global rules.
Smith identified the UK as a potential leader in global AI safety governance. The UK AI Security Institute (AISI), established in 2023, conducts independent safety assessments of new models from frontier labs. Despite recent pressure from the US government to restrict access to these models, the AISI confirmed this week that it maintains the ability to test the latest AI systems.
The Microsoft executive suggested that the UK's position is unique because it is not viewed as an aggressive regulator, unlike the EU. This status allows the UK to act as a bridge in international negotiations. Smith believes this diplomatic advantage is crucial for fostering global consensus on safety standards.
In October, US President Donald Trump convened executives from companies including OpenAI, Anthropic, Nvidia, AMD, SpaceX, Meta, and Google to discuss AI safety. The participants signed a voluntary agreement to cooperate with independent audits. Smith welcomed the move but argued that voluntary self-regulation is insufficient to address risks from non-compliant actors.
Smith proposed several mandatory safety measures, including a legal requirement to report breaches caused by AI agents. He also advocated for mandatory external evaluations prior to product releases and the implementation of emergency stop mechanisms. These measures aim to ensure human control over advanced AI systems.
