Bill Gates and Microsoft president Brad Smith say AI self-regulation is no longer enough; Smith: 'your washing machine is subject to more safety rules than AI'

Why it matters

【Key points】

Over the same weekend, Microsoft co-founder Bill Gates and Microsoft president Brad Smith both said AI industry self-regulation is no longer enough.

On CNN's State of the Union (Oct 11), Gates said the real near-term danger is not AI getting out of control, which is years away, but bad actors using the latest models with safeguards removed: "over the next year, primarily in the cyber area, we are going to have a lot of negative surprises," because extortion and market or betting disruption already have a business model. He expects the industry to "move beyond self-governance" within a year, with government safety bodies like those for planes and medicines.

On the BBC's Big Boss Interview (released Oct 11), Smith said: "Right now, your washing machine is subject to more safety rules than AI." He proposed an emergency brake for AI running autonomous devices or critical infrastructure, at both the model-company and cloud layer; a legal duty to tell victims when rogue AI agents breach their systems; and, most important, deciding whether government evaluators must sign off frontier models before release. He named the UK AI Security Institute as best placed to lead.

【Verifiable】

CNN transcript: https://transcripts.cnn.com/show/sotu/date/2026-10-11/segment/01

BBC article: https://www.bbc.com/news/articles/c6rmy7zmvx3lo

Smith called the White House "morally binding" lab accord (e/815) "a necessary step" Microsoft would sign, "but I don't think it's sufficient any longer." About 660 X posts in 24h, still 30-50 per hour.

【Our read】

1. Both echo the UK package (e/827) and US mandatory incident reporting (e/795), but Smith puts pre-release government sign-off at the top, further than either government has gone.

2. A brake at the cloud layer implies cloud providers, Microsoft included, share shutdown duty; watch whether Microsoft's own policies follow.

3. Gates's one-year cyber warning is a forecast, not data, and names no model or incident.