Microsoft chief: Washing machines face more safety rules than AI

Voluntary pledges from tech companies are not enough to keep AI safe. That is according to Brad Smith, president of Microsoft, in a BBC interview reported on Oct. 11-12, 2026. ITV News reported on his views as early as Oct. 8.

Smith compared AI with everyday products: "Right now, your washing machine is subject to more safety rules than AI." He also used milk as an example. When you buy a carton of milk, you don't have to worry about whether it is safe, because regulations set a safety floor.

Voluntary agreements are a good step, but not enough

According to Smith, voluntary safety agreements are a good and necessary step, but they are no longer sufficient. His reasoning is that the responsible companies will keep their voluntary pledges. It is the less responsible companies that are the concern.

Smith refers to a one-page voluntary AI safety agreement from a White House summit. According to Coin Alert News, signatories include OpenAI, Anthropic, Nvidia, Meta, Google and Elon Musk's X. Microsoft has not signed, but is willing to. The sources describe the agreement differently, however: One refers to a pledge of independent audits, the other to a general safety agreement.

What does he propose?

Smith proposes, among other things:

- mandatory global safety standards for AI - legally required reporting of failures in AI models and of incidents - that companies must give notice when rogue AI agents break into systems, and that victims must be notified - independent review of powerful models before they are released - emergency brakes, also called "kill switches," that can pause AI systems or shut them down

These are policy recommendations. They have not been introduced as new requirements at Microsoft, as legislation or as binding industry standards.

Pointing to the UK

Smith points to the UK as the country best positioned to set global AI standards, partly because of its AI Security Institute. According to Coin Alert News, the institute tests advanced models from US labs before they are put into use.

The UK is also preparing a cybersecurity and resilience bill that would require reporting of serious AI incidents. The country also takes over the G20 presidency in 2027, where AI is set to be a key theme.

According to The News International, Smith's position departs from President Trump's line. Trump favors industry self-regulation to keep American companies competitive with Chinese ones.