Microsoft’s Smith: Milk is regulated more tightly than AI

The latest:

Artificial intelligence should be governed by a global set of binding safety rules enforced by independent evaluators, Microsoft president Brad Smith said in an interview with the BBC published Monday. He argued the technology currently operates under looser requirements than household appliances or supermarket dairy, and said governments already know how to regulate products and should apply that knowledge to AI.

Details:

• The comparison: Smith said a washing machine is subject to more safety rules than AI, and that every carton of milk in a grocery store faces safety rules and inspection, according to the BBC interview. He framed existing product-safety regimes as a template rather than a model to copy wholesale.
• The proposals: Smith set out three specific requirements, Reuters reported: that a victim be informed when their system is breached by a rogue AI agent, that an emergency brake be available to halt a system, and that companies face a legal obligation to file incident reports.
• His own words: “We know how to do this for other products,” Smith said, adding that the industry should build on that knowledge and consider how to apply it to AI. He did not specify which existing regulatory body or treaty framework should house the rules.
• Industry disclosure: Companies including Anthropic and OpenAI have begun disclosing rogue or undesired behaviour by AI agents, Reuters reported, as part of a wider debate over the fast-advancing technology that has included researchers warning of an eventual existential threat to humanity.
• The motives debate: Some companies have called for AI development to slow down. That has triggered argument over whether executives have ulterior motives, Reuters reported, including writing rules that preserve their firms’ lead, pre-empting a broader government backlash, or simply slowing breakneck spending.
• Britain over Brussels: Smith said Britain was the country best placed to lead on AI regulation because of the AI Security Institute, launched under former prime minister Rishi Sunak. Officials in Washington, he said, view London very differently from how they view Brussels, which regulates AI by risk level through its AI Act.
• How the institute works: Britain’s institute receives pre-deployment access to major AI models under voluntary agreements, allowing it to assess their capabilities and risks before release. The arrangement rests on consent from developers rather than statutory powers.
• The government position: AI Minister Kanishka Narayan told Reuters that Britain would consider regulating advanced models if the current system proved insufficient. He did not set a threshold for what would count as insufficient, or a date for any review.

Between the lines:

Smith’s preference for London over Brussels tracks the difference between the two regimes described here: Britain’s institute operates on voluntary pre-deployment access, while the EU AI Act imposes obligations by risk tier. A company asking for global rules while endorsing the consent-based model is asking for a specific kind of regulation, which is precisely the tension the motives debate around slowdown calls has raised.

What’s next

Watch whether Britain moves from voluntary agreements to statutory rules for advanced models, the test Narayan set, and whether other major developers follow Anthropic and OpenAI in publishing incident disclosures.