- Lead. Microsoft president Brad Smith told the BBC on October 12 that AI systems currently face lighter regulatory scrutiny than a carton of milk, and called for a global safety framework with mandatory incident reporting and an emergency brake for developers.
- Fact. Smith singled out Britain’s AI Security Institute — established under former Prime Minister Rishi Sunak — as a model for how governments can conduct pre-deployment model evaluations, and said Washington views London’s approach more favourably than the EU’s risk-based framework.
- Stake. The intervention by one of the industry’s most senior figures arrives as questions about AI company disclosures intensify: OpenAI’s $20 billion revenue accounting gap last week illustrated how loosely the sector’s commercial claims remain scrutinised.
Brad Smith, president of Microsoft, gave an interview to the BBC published Monday in which he argued that the absence of safety regulation for advanced AI represents a governance failure comparable to selling uninspected food. “Every carton of milk that you see in the grocery store is subject to safety rules and inspection,” Smith said. “We know how to do this for other products. We should build on that knowledge and think about how to apply it to AI.” He made the same point about washing machines, which he said are more tightly regulated than AI applications that can operate autonomously at scale, according to Fox Business.
What Smith Proposed
Smith outlined four specific regulatory mechanisms: a global set of AI safety standards that governments would jointly adopt; a notification mechanism for individuals whose systems are compromised by an autonomous AI agent; an emergency brake that developers and regulators could invoke to halt a model’s operation if unexpected harm emerges; and mandatory incident reports when AI causes adverse outcomes, analogous to aviation or pharmaceutical reporting requirements. He cited those industries as domains where comparable frameworks already exist and function effectively.
Britain’s Position and the US View
Smith said the United Kingdom is well-positioned to lead on AI regulation because of its AI Security Institute, which operates under voluntary agreements to receive pre-deployment access to frontier models. UK AI Minister Kanishka Narayan told Reuters separately that Britain would consider moving to mandatory regulation of advanced models if the voluntary system proved insufficient. Smith also noted that the US government views London’s approach quite differently from the EU’s AI Act, which applies risk-based categories and carries significant compliance requirements. The EU framework has faced criticism from parts of the US technology industry as overly prescriptive; Smith’s remarks suggest Microsoft would prefer Britain’s safety-institute model to expand globally.
The Industry Backdrop
Smith’s remarks are notable because Microsoft is itself one of the world’s largest AI developers, with its Azure platform hosting OpenAI’s models and its Copilot system embedded across products used by hundreds of millions of people. OpenAI’s Sam Altman and Anthropic’s Dario Amodei have made similar calls for third-party evaluators with employee-like access to AI development processes. Britain’s AI Security Institute already receives that access on a voluntary basis, making it the closest existing approximation to what major industry players say they want — a framework shaped partly by the companies it would govern.