Microsoft president: Milk has more safety rules than AI
Microsoft's president wants mandatory emergency brakes and independent testing for AI, warning that voluntary pledges are no longer enough.
Policy · Source: Superpower Daily (BBC Big Boss Interview)
What happened
Microsoft president Brad Smith is demanding strict global rules for artificial intelligence. He says AI is currently less regulated than a carton of milk or a washing machine. In recent interviews, he argued that voluntary safety pledges cannot stop irresponsible competitors. He wants governments to step in now before the technology spirals out of control.
Smith proposed three specific layers of regulation to handle the risks. First, independent government evaluators must test select AI systems before they are released to the public. Second, companies must build emergency brakes or kill switches to pause rogue systems. Third, developers must face legal requirements to write incident reports and notify victims if an AI agent breaches their systems.
The Microsoft executive pointed to the UK as a potential leader for global standards. He highlighted the British AI Security Institute as a model for testing frontier technology. Smith noted that Washington might be more receptive to rules coming from London rather than Brussels. He acknowledged the recent voluntary White House agreement signed by tech executives, but insisted that promises from willing companies are not sufficient.
Key facts
- Brad Smith — Microsoft president calling for global AI safety rules
- AI Security Institute — UK body Smith cited to help set international standards
- Donald Trump — US President who prefers industry self-regulation for AI
Why it matters
If these proposals become law, building AI products will look a lot more like manufacturing physical goods. You will not be able to just ship a model and patch it later. Startups and infrastructure providers will need to build mandatory kill switches and compliance reporting directly into their tech stacks. The era of moving fast and breaking things in AI is facing a massive corporate pushback. The focus is shifting entirely toward accountability, safety testing, and strict disclosure requirements.
This shifts the competitive landscape heavily in favor of massive incumbents. Microsoft can easily afford to hire armies of compliance officers and pay for independent government evaluations. Small teams and open-source developers will struggle to shoulder the legal liability of mandatory incident reporting and victim notification. Regulation designed to stop rogue agents will inevitably act as a moat for the biggest tech giants. The open question is whether these voluntary pledges will actually turn into shared requirements.
For builders
Mandatory kill switches for AI agents
Engineers will need to design emergency brakes that can pause or slow down autonomous systems. Infrastructure providers who build these intervention mechanisms as a service will get paid. Startups that ignore this architecture risk getting shut out of future compliance frameworks.
Legal liability for rogue AI breaches
Smith wants mandatory incident reports and victim notifications when an AI agent accesses systems in unintended ways. Founders building autonomous agents will carry the legal risk if their product goes rogue. You will pay for the legal fallout, making robust testing a survival requirement.
Pre-release government safety evaluations
Independent evaluators would need to vet selected AI technology before it reaches users. This adds massive friction and cost to product launch cycles. Big tech pays the compliance fees easily, while startups lose their speed advantage.
My take
I find it entirely predictable that the company selling the most AI is begging for regulation. When Brad Smith says AI needs more rules than milk, he is really saying Microsoft wants a regulatory moat. We are watching the big players use safety concerns to pull up the ladder behind them.
Original reporting: Superpower Daily (BBC Big Boss Interview). This is my rewrite and opinion.