Microsoft president calls for AI incident reports, emergency brakes and outside checks

Microsoft president Brad Smith wants AI firms to face safety rules that go beyond voluntary promises. His proposals include reporting serious incidents, telling people affected by rogue AI agents and giving independent evaluators a role before products are released.

Microsoft AI Watch analysis

What happened

In a BBC interview published on 11 October, Smith argued that AI is subject to fewer safety rules than everyday products such as milk and washing machines. He called for “emergency brakes” to pause or slow an AI system if it gets out of control, alongside legal duties to file incident reports and notify people affected by rogue AI agents.

Smith also suggested that independent government evaluators should assess new AI technology before release. He said the UK was well placed to lead international rule-making, pointing to its AI Security Institute. The BBC’s interview report says Smith welcomed a recent voluntary commitment by technology companies to work with independent auditors, but argued it was not enough on its own.

Why it matters

These proposals turn the broad demand for “AI safety” into questions of who checks systems, what companies must disclose and what happens when something goes wrong. Incident reporting and telling affected people would make failures harder to keep quietly in the filing cabinet; external assessment would put at least some scrutiny outside the companies building the systems.

Smith’s comments also expose a real divide over whether voluntary industry pledges can do the job. He argues that responsible companies may comply, while less responsible ones are the problem. The interview records his position, not an agreed policy or proof that current AI systems have escaped human control.

Our read

“Emergency brakes” makes for a vivid phrase, but the useful test is the rulebook behind it: what triggers a pause, who decides, and what must be reported afterwards? Smith has put concrete ideas on the table. Legislators still have the harder task of making them enforceable without turning safety checks into a rubber stamp.

What to watch

• Whether the UK or other governments propose mandatory AI incident reporting.
• Who would conduct independent assessments, and when they would happen.
• Whether firms’ voluntary audit commitments become binding rules.
• Whether any proposed notification duty defines who counts as affected by an AI incident.

Discussion spark: Should governments require independent safety checks and incident reports for powerful AI systems, or can voluntary industry audits provide meaningful accountability?

not affiliated with or endorsed by Microsoft