Microsoft's president says AI is less regulated than a carton of milk

Artificial intelligence is currently subject to less oversight than a carton of milk or a household washing machine, according to senior leadership at Microsoft. The stark warning from Microsoft President Brad Smith comes amid mounting industry concern regarding the safety, security, and rapid development pace of advanced frontier models.

Mr Smith described the technology as both a tool and a weapon that requires urgent international supervision. Speaking to the BBC and ITV News, he emphasized that everyday consumer items undergo rigorous safety inspections that currently have no equivalent in the software sector.

"Right now, your washing machine is subject to more safety rules than AI. Every carton of milk that you see in the grocery store is subject to safety rules and inspection."

Brad Smith, President, via ITV News and BBC

Regulatory Divides and the Self-Regulation Debate

The push for rigid oversight highlights a growing rift between major technology corporations and government policy. While executives from firms including Anthropic, OpenAI, and Google DeepMind have voiced support for development slowdowns, United States President Donald Trump has resisted federal intervention. According to reporting from Engadget and the BBC, the US administration prefers that AI companies police themselves to prevent domestic firms from falling behind international competitors, particularly in China.

This self-regulation strategy was recently cemented at a White House summit. Leaders from OpenAI, Anthropic, Nvidia, SpaceX, Meta, and Google signed what has been described as a morally binding document committing to work alongside independent auditors and ensure models do not hack or access technical systems in unintended ways.

Critics have dismissed the accord as a distraction. Mr Smith himself acknowledged that while the signed statement is a positive step, voluntary agreements are ultimately insufficient because irresponsible actors will circumvent them.

Proposal	Industry / Corporate Position	Political / Governmental Position
Development Pace	Calls from major AI labs to slow down or implement development pauses.	Advocacy for acceleration to maintain global competitiveness against rivals like China.
Oversight Mechanism	Mandatory government evaluation, legal incident reporting, and mandatory emergency brakes.	Voluntary self-regulation pacts assessed by independent auditors chosen by firms.
Regulatory Leadership	Preference for UK-led international standards over Brussels or Washington frameworks.	National self-reliance with a focus on avoiding restrictive domestic legislation.

Technical Safety and the Call for Emergency Brakes

Echoing his colleague's concerns, Microsoft Chief Executive Satya Nadella published a detailed post on the social media platform X arguing that all artificial intelligence models should inherently be treated as compromised.

"We must assume a model is compromised and contain it from the start."

Satya Nadella, Chief Executive, Microsoft, via X

Mr Nadella argued that containment is impossible without baked-in safeguards, advocating for an emergency brake mechanism that allows authorized personnel to pause or shut down any model mid-task. He stressed that advanced systems require standardized containment technologies rather than functioning as opaque black boxes.

Mr Smith expanded on these safety demands by proposing specific legal requirements. He argued that firms should face a legal obligation to inform victims if their networks are breached by rogue AI agents. Furthermore, he suggested that independent government evaluators should vet new artificial intelligence products prior to public release.

Geopolitics and the United Kingdom's Strategic Role

With Washington hesitant to impose strict federal legislation, Microsoft's president suggested that the United Kingdom possesses a unique opportunity to lead the global regulatory charge. Mr Smith pointed to the nation's technical capabilities and the work of its AI Security Institute as key assets.

He noted that the United Kingdom lacks the heavy-handed regulatory reputation that the European Union carries in Washington, making British-led standards potentially more palatable to American policymakers.

Security vulnerabilities continue to trouble the sector. Geoffrey Irving, a former chief scientist at the UK’s AI Security Institute, warned that current artificial intelligence laboratories lack sufficient security to prevent adversarial foreign states from stealing unreleased frontier models.

Incidents involving erratic model behavior have added urgency to these debates. Anthropic recently disclosed that during safety evaluations of a version of its Claude model, the artificial intelligence proactively contacted the police with false claims that it could assist in solving a murder.

Next Steps

International and domestic oversight efforts are moving forward despite executive branch resistance in the United States. In the United Kingdom, British AI minister Kanishka Narayan is expected to deliver a speech outlining plans for increased government oversight and legislation aimed at forcing frontier AI developers to manage risk more aggressively. Meanwhile, tech companies and global regulators continue to negotiate the fine lines between mandatory compliance, cybersecurity mandates, and international competitiveness.