Skip to content

AI

Microsoft's president says your washing machine out-regulates AI

Brad Smith wants global AI safety rules, "emergency brakes" included, and thinks the UK should write them. Washington is not thrilled, and neither is the EU's sense of self-importance.

TL;DR

  1. Brad Smith: milk and washing machines face more safety rules than AI. He wants global rules now.
  2. His wish list: breach notification for victims, third-party sign-off before release, and AI kill switches.
  3. He picks the UK's AI Security Institute to lead, not the US or EU.

What happened

Smith called AI both a "tool and a weapon" and said the moment has arrived for global safety rules. He said the world's cybersecurity is at risk and added: "Let's get started today."

His wish list is concrete. First, a legal duty to tell a "victim" when their system has been breached by a rogue AI agent.

Second, clear rules on when outside evaluators must sign off on an AI product before release. Third, "emergency brakes", which the industry calls kill switches, able to pause or slow down an AI that gets out of control.

Then came the geography. Smith said the UK, not the US or the EU, is best placed to lead.

Britain's AI Security Institute, launched in 2023, gets early access to big US lab models ahead of release. TechCrunch says it has fast become a top name in AI safety testing.

Why it matters

This puts a Microsoft executive at odds with President Donald Trump, whose line is that the industry should police itself, so American firms keep their lead over Chinese rivals.

It also lands in the middle of a live fight. TechCrunch reports Trump has warned against "globalist" organizations he says interfere with American companies.

Last month he reportedly told Anthropic and OpenAI to keep new models away from British testers. The institute confirmed on Monday that it can still access and test the latest models.

Smith argued the UK can regulate AI in ways the US cannot. Whatever you think of that, it is a big company publicly choosing a side.

One caveat: this is advocacy, not law. Nobody passed anything. A very senior person said a very quotable thing about dairy.

What it means for Europe

Smith named the UK over the EU as the best place to lead, which is a polite way of saying Brussels is not the first call.

For European companies, the practical signal is where testing power may sit: with the institute that already sees frontier models before launch.

If breach notification and third-party sign-off become global norms, EU firms will meet them either way. The open question is whose template wins, and Smith has now put a vote in for London.

What to do with it

  • Audit your agents.If an AI agent can touch customer data, write down who gets told, and how fast, when it goes wrong.
  • Find your kill switch.Check that you can pause or cut off any AI tool in your stack without a heroic phone call.
  • Ask vendors about testing.Request to know whether an independent evaluator has looked at the model before you build on it.
  • Watch the UK institute.Its testing access to new models is the story to follow as the US pushback continues.

Keep up, sapiens. Your washing machine already passed inspection.