NeuralStackArtificial intelligence is less regulated than milk and washing machines, Microsoft president Brad Smith has warned, calling for global safety rules to govern the technology. Speaking to the BBC’s Big Boss Interview podcast, Smith described AI as both a “tool and a weapon” and said regulation should begin immediately.
Smith said companies should be legally required to notify victims when a rogue AI agent breaches their systems. He also called for firms to produce incident reports when AI systems go out of control, creating a formal record of failures and their consequences.
Another part of his proposal is the introduction of emergency brakes, often called kill switches, that could pause or slow an AI system if it began behaving dangerously. Smith also suggested that independent government evaluators should assess new AI products before release.
“Right now, your washing machine is subject to more safety rules than AI,” Smith said. He compared AI oversight with the safety rules and inspections applied to cartons of milk, arguing that regulators already know how to assess other products and should adapt that experience to artificial intelligence.
Smith’s position differs from President Donald Trump’s approach. Trump has argued that the sector should regulate itself so US companies do not lose ground to Chinese rivals, while Smith believes voluntary commitments are no longer enough to manage irresponsible developers.
At the end of last month, leaders from OpenAI, Anthropic, Nvidia, SpaceX, Meta and Google signed a “morally binding” document at the White House. The companies agreed to work with independent auditors to assess whether their AI systems operate as intended and avoid unintended access to technical systems, but Smith said the statement was necessary without being sufficient.
Smith said the UK is well placed to lead international efforts because of its AI Security Institute, established under former prime minister Rishi Sunak’s government. He suggested Washington may be more receptive to standards led by the UK than rules associated with the European Union, because the UK is viewed differently in the US capital.
Why this matters
Smith’s comments sharpen a growing divide over how AI should be governed: through voluntary industry promises or enforceable obligations. Mandatory reporting, notification and independent evaluation would give governments and affected users more visibility when systems fail, while emergency brakes could provide a practical way to contain dangerous behavior.
He said the moment had arrived for governments to act calmly but firmly, particularly as AI agents become capable of operating beyond direct human control and powerful models create cybersecurity risks. Whether the UK can turn its technical expertise into broadly accepted global rules will determine how far Smith’s proposal moves beyond another industry pledge.