Microsoft’s Brad Smith Says the UK Is Best Positioned to Lead Global AI Rules

Introduction

The debate over AI governance is moving beyond broad promises of responsible innovation and toward specific obligations for companies developing frontier systems. In a BBC podcast appearance, Microsoft President Brad Smith argued that AI should be treated as both a tool and a potential weapon, with implications for global cybersecurity. In his view, the time has arrived for international safety rules, and the United Kingdom may be better positioned than either the United States or the European Union to help drive them.

Key points

• Voluntary commitments have limits. Several major technology companies recently agreed to work with independent auditors under a document described as “morally binding.” Smith called that a constructive step, but said it cannot replace enforceable regulation. Companies that already behave responsibly are the most likely to comply voluntarily; the greater concern is the group that does not.
• The UK has a testing platform. The UK’s AI Safety Institute, established in 2023, has been given access to models from leading US laboratories before deployment. That access has helped it become an important participant in frontier-model safety testing.
• Smith wants operational requirements. His proposals include a legal duty to notify an appropriate organization when a system is compromised by rogue AI agents. He also spoke favorably about independent assessments before release and technical “kill switches” that could slow or pause a system if humans lose control.
• London may be a less confrontational intermediary. Smith said policymakers in Washington do not view London in the same way they view Brussels. He sees that political distance as a potential advantage for bringing US companies, European authorities, and other governments into the same conversation.

Why it matters

Smith’s comments point to a shift in the policy debate. The question is no longer only whether companies should act responsibly, but which safety actions should be mandatory and who should have the authority to verify them. If pre-release evaluations, incident reporting, and emergency intervention become legal requirements, they could turn safety testing into a standard part of launching an advanced AI product rather than an optional corporate practice.

The proposal also highlights the UK’s attempt to build influence through technical evaluation rather than through the market size or regulatory reach associated with the US and EU. That role is not guaranteed. The source report describes recent uncertainty over whether British testers would continue to receive access to new models amid political pressure from Washington, although the AI Safety Institute said it could still access and test the latest systems.

The broader lesson is that self-regulation and public regulation may work best as complementary layers. Labs can contribute testing methods and technical expertise, while governments establish minimum duties that apply even to companies unwilling to police themselves. The difficult balance will be designing rules that can respond to serious loss-of-control scenarios without creating a blanket barrier to useful AI development.

Source: TechCrunch AI