Microsoft president Brad Smith has called for urgent global AI regulation, saying Britain is best placed to lead it.
Smith said AI should be subject to global rules overseen by independent evaluators who could assess new developments. He identified Britain as the country most capable of taking the lead because it has an AI Security Institute.
I hope that's the role we can play as a company. I hope that's the role the UK can play as a country. [W]hen you look at models that are so powerful that they're putting the world's cybersecurity at risk, I would say: the moment has arrived.
Smith said artificial intelligence is subject to fewer safety requirements than milk, even though it can attack companies and governments. Systems could threaten global cybersecurity and therefore required safeguards that applied across national borders.
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.
His appeal comes during a wider argument over how to control increasingly capable systems without stopping development indefinitely. Researchers have warned of an eventual existential threat to humanity, while some companies have called for development to slow. That has prompted debate over whether executives are seeking to protect their firms' lead, avoid a wider government backlash or restrain spending.
Under Smith's proposal, a victim would have to be informed if a rogue AI agent breached their system. He also wants an emergency brake available for AI systems and a legal duty on companies to write incident reports.
We know how to do this for other products.
We should build on that knowledge and think about how to apply it to AI.
The proposals draw on existing approaches to product safety, but the pack does not establish which governments would adopt them or how they would enforce common rules. The question of whether Britain will legislate, and when, remains open.
Britain's voluntary model faces a test
Britain's AI Security Institute was launched under former prime minister Rishi Sunak. Voluntary agreements with major AI model developers give the institute access before deployment, allowing it to examine capabilities and risks.
- Under Rishi SunakInstitute launched
Britain created its AI Security Institute.
- Current systemVoluntary testing
The institute receives pre-deployment access to major AI models under voluntary agreements.
- 12 October 2026Ministerial speech
Kanishka Narayan is expected to call for stronger safeguards and discuss possible legislation.
The institute is central to Smith's view that Britain can bridge different approaches to oversight. AI Minister Kanishka Narayan has told Reuters that the government would consider regulating advanced models if the existing system proved insufficient.
Britain and the European Union
Britain
Voluntary model
- AI Security Institute
- Pre-deployment access to major models
- Possible legislation if the current system proves insufficient
European Union
AI Act
- Regulation based on the level of risk
The European Union already regulates AI under its AI Act according to the level of risk. Britain has not decided whether to impose tougher requirements on advanced models, so its current arrangements remain voluntary.
The United States has taken a different path. Smith said the lack of US action created a unique opening for Britain, while US President Donald Trump has said AI companies should regulate themselves rather than face government rules.
In Washington DC, people don't look at the UK and they don't look at London the way they look at Brussels — that's a strength,
At Trump's so-called Super Intelligence summit, Google's Sundar Pichai and Anthropic's Dario Amodei joined NVIDIA CEO Jensen Huang, Elon Musk, OpenAI President Greg Brockman and Mark Zuckerberg to sign a pact. The pact said its principles included self-regulation, safeguards intended to stop pioneer models from going rogue and independent auditors to assess models.
Critics of the pact said it allowed technology companies to decide for themselves what safety meant and that the agreement should be ignored as a distraction.
Security concerns sharpen the debate
Calls for stronger controls have grown among experts, politicians and regulators. Anthropic, Google DeepMind and OpenAI have also warned that the pace of development should slow.
Recent concerns include deceptive behaviour by AI systems and the possibility that hostile states could steal powerful models before they are released. Anthropic said that, during testing of a version of its Claude model, the system falsely told police it could help solve a murder after contacting them.
not have security that would prevent an adversarial country from stealing the models
Irving's warning focused on the protection available to companies against an adversarial country taking models. Experts have also said enemy states could steal unreleased, powerful systems.
Smith said AI was both a tool and a weapon. Systems need mechanisms that can stop or slow them if their behaviour becomes dangerous.
We can’t treat super intelligence as a set of nested black boxes and simply accept or reject its recommendations, answers, and actions.
We must build contained systems whose behaviour we can observe, limits we can test, and actions we can always contain.
Nadella called for an emergency brake on Saturday. Smith has referred to such controls as emergency brakes or kill switches, which could pause or slow a system that moved out of control.
Microsoft operates AI infrastructure, holds a stake in OpenAI, the owner of ChatGPT, and is led by Nadella. The company therefore has a direct role in the technology and in the debate over how it should be governed.
The immediate test for Britain is whether its voluntary arrangements produce legislation and stronger oversight. It remains unclear whether the UK government will regulate advanced models, what worldwide safety rules will be adopted or which governments will enforce them.
On 12 October 2026, Kanishka Narayan is expected to call for stronger frontier-AI safeguards and discuss possible legislation.
