Microsoft’s Brad Smith Calls for Urgent AI Regulation, Backs UK Leadership

TL;DR

• Microsoft president Brad Smith urged governments to introduce enforceable global AI safety rules.

• He argued that everyday products such as milk and washing machines face more oversight than AI developers.

• Smith said the UK is well positioned to lead international regulation through its AI Security Institute.

• He welcomed a voluntary White House safety agreement but said it cannot replace binding requirements.

Microsoft president Brad Smith has called for urgent AI regulation, warning that increasingly capable systems pose cybersecurity risks that voluntary industry commitments cannot adequately address.

Speaking on a recent BBC podcast, Smith described AI as both a “weapon and a tool” and argued that the time had come to establish global safety requirements.

He compared oversight of AI developers with regulation of everyday consumer products, saying milk and washing machines face stronger controls.

Smith Says the UK Can Lead Global AI Safety Rules

Smith identified the UK as particularly well placed to coordinate international AI regulation.

Britain’s AI Security Institute, launched in 2023, tests major models and receives access from U.S. frontier laboratories before deployment.

Its role has faced uncertainty following reports that President Donald Trump asked OpenAI and Anthropic to withhold new models from British testers pending U.S. review.

The institute nevertheless confirmed Monday that it could access and test the latest models.

Smith argued that Washington views London differently from Brussels, giving Britain a diplomatic advantage in building support for international safeguards.

Voluntary Commitments Leave Enforcement Gaps

Last month, Trump hosted executives from companies including OpenAI, Anthropic, Nvidia, AMD, SpaceX, Meta and Google for discussions about AI safety.

The participants signed a document described as “morally binding,” agreeing to work with independent auditors to assess model safety and capabilities.

Smith welcomed the agreement as progress but said it fell short of the regulation required.

His concern is that voluntary commitments attract responsible companies while leaving less accountable developers free to avoid them.

Binding rules, he argued, are needed to address companies that would otherwise decline to participate.

Proposed Rules Cover Breaches, Evaluations and Emergency Controls

Smith outlined several safeguards he would like governments to establish.

One would create a legal obligation to notify an organization if rogue AI agents breached its systems.

He also expressed support for outside evaluators approving AI products before release and for emergency controls that could pause or slow systems if people lose control.

The proposals address both prevention and response: assessing risks before deployment, notifying affected organizations, and retaining a way to intervene during an incident.

Industry Leaders Face Growing Questions About Agent Safety

Smith’s remarks come amid calls from executives including Anthropic’s Dario Amodei and OpenAI’s Sam Altman for greater caution in AI development following incidents involving agents acting outside intended boundaries.

The debate is becoming more consequential as AI systems gain tools and the ability to complete tasks with less direct supervision.

Smith urged companies and governments to build consensus calmly but firmly, saying he hoped Microsoft and the UK could help bring participants together.