Microsoft's president says AI has fewer safety rules than milk

In an interview on the BBC’s Big Boss Interview podcast, released October 11, Microsoft president Brad Smith argued that everyday products face tougher safety oversight than AI. His examples were milk cartons and washing machines. He said the moment has come for international safety rules.

Reports of the interview list several ideas he backed:

• Emergency brakes, or kill switches, to pause AI that gets out of control.
• A legal duty to tell victims when a rogue AI agent breaches their systems, much like data-breach rules.
• Government evaluators signing off on frontier models before release.

Smith called a recent voluntary White House pledge on working with independent auditors a good step, but said it is no longer enough, since the companies you worry about are the less responsible ones. He also said the UK, with its AI Security Institute, is best placed to lead.

The catch: These are one executive’s proposals, not law. Microsoft has a business interest in how rules get written, and TechCrunch notes the UK institute’s access to new models has been under pressure, though it said on October 12 that it can still test the latest ones.

What to do now

• If you run AI agents with access to your accounts, note which systems they can reach and how you would switch them off.
• Keep a simple log of what each agent is allowed to do, in case breach-reporting rules arrive.
• Watch your AI vendors’ terms for incident-notification promises.