“Right now, your washing machine is subject to more safety rules than AI.” That quote, from Microsoft President Brad Smith in a BBC Big Boss Interview published 11 October 2026, captures a regulatory argument now driving a serious split between the AI industry, the US administration, and global policymakers.
Smith’s argument is pointed: the systems now influencing critical infrastructure, cybersecurity, and autonomous devices face fewer mandatory safety requirements than everyday consumer products, in his assessment. He described AI as both a “tool and a weapon,” and said industry self-regulation is “not sufficient any longer.”
The Regulatory Gap
Voluntary promises with no stated penalties are not the same as enforceable safety rules.
A White House agreement involving major technology companies established independent external auditors and internal controls. Those controls are intended to prevent AI systems from accessing unintended networks or behaving outside their designed limits. Reuters reported the agreement was described as “morally binding” and contained no stated penalties for noncompliance.
Smith welcomed that agreement as a necessary step but said it is no longer enough. Responsible companies already follow voluntary expectations; the problem, he argued, is that less responsible ones pose the greater risk and face no legal obligation to comply.
To close that gap, Smith outlined four specific proposals:
- Emergency brakes (commonly called kill switches) built into both the companies operating AI models and the cloud infrastructure hosting them, so dangerous behavior can be paused or stopped
- Mandatory incident reporting, requiring companies to notify authorities when an AI agent escapes its intended limits or accesses systems it was not designed to reach
- Victim notification after rogue AI incidents, modeled on existing data-breach disclosure requirements, so affected people and organizations are informed
- Independent pre-release evaluation of the most advanced AI models, known as frontier models, by government evaluators before those systems are released; Smith described this as his most significant proposal
The US–UK–EU Triangle
Smith’s call for binding global rules places him directly at odds with President Trump, who has promoted self-regulation as the approach most likely to keep US firms competitive against China. The voluntary-safeguards position reflects a view that mandatory oversight risks slowing American AI development at a strategically critical moment.
Smith’s counterargument is practical: voluntary rules only govern the companies already behaving responsibly. They offer no leverage over those that won’t.
On the question of where enforceable global standards might come from, Smith pointed to the UK. He argued that the UK AI Security Institute (formerly the AI Safety Institute) has the technical capacity to lead international safety evaluation. Washington views London differently from Brussels, he noted, which makes the UK a credible intermediary between US and European regulatory approaches.
There is a significant constraint on that role. The UK institute currently operates through voluntary access arrangements with AI companies rather than mandatory powers, meaning it can test models only when developers agree to share them.
The central policy question remains unresolved: whether frontier models will face mandatory government evaluation before release or remain subject to voluntary testing. That choice will shape how the AI systems already affecting your daily life are built, deployed, and held to account.
