Microsoft, OpenAI executives call for AI safety laws, says warnings alone are not enough

Microsoft Vice Chair and President Brad Smith on Wednesday called for new artificial intelligence safety laws, saying tech companies must prevent harm rather than simply warn of risks.

Speaking at the International AI Summit 2026 in Ireland, Smith said AI capabilities had advanced faster than protections.

“The world here in Europe or anywhere does not expect our industry simply to warn about what could go wrong. They expect us to prevent it. They expect us to take action,” he said.

“And if we need to slow down, then it is up to each company and each leader to slow down.”

Describing 2026 as “a sobering year for AI safety,” Smith cited reports of rogue AI agents hacking networks and AI worms targeting connected homes.

“There are more safety rules for a carton of milk than for AI. There are more safety rules for your oven than for AI,” he said.

He urged the industry to overcome resistance to regulation and to back incident reporting, notification of victims of AI-enabled breaches, and independent evaluations with findings shared with governments and customers.

“But we need law. We need regulation. It needs to be sensible. It needs to be balanced. It needs to take into account innovation,” Smith said.

“We need to recognize that the power of AI has moved faster than the safety for it. Safety needs to catch up.”

'We need to make sure the defenders have the tools'

Also speaking at the event, OpenAI Chief Financial Officer Sarah Friar supported cooperation among leading AI developers, saying the company had called for global regulation.

Friar said OpenAI is withholding an unreleased model while doing further safety and alignment work, but cautioned against slowing development so sharply that cyber defenders lose ground.

“So the folks with malicious intent have the tools. So we need to make sure the defenders have the tools,” she said.

She said OpenAI had used AI to identify vulnerabilities in its own infrastructure.

“We found the vulnerabilities, we patched them at machine speed,” Friar said.

“Selling fear is not a good way to come to the world. We have to sell solutions and solutions that feel transparent, open, honest, and so on.”