Key points
- Microsoft President Brad Smith calls for urgent global AI safety rules, stating current oversight is weaker than that for consumer goods.
- The UK AI Security Institute retains access to test latest models from US labs despite political pressure to withhold them.
- Tech leaders signed a voluntary safety document, which Smith views as insufficient compared to mandatory legal obligations.
Microsoft President Brad Smith has called for immediate global regulation of artificial intelligence, arguing that current oversight is inadequate. He told the BBC that consumer products such as milk and washing machines face stricter scrutiny than companies developing AI systems.
What happened
During a recent BBC podcast appearance, Microsoft President Brad Smith described AI as both a weapon and a tool that jeopardises global cybersecurity. He stated that the moment has arrived to institute global safety rules, arguing that the United Kingdom is best placed to lead this effort rather than the US or European Union. Smith highlighted the role of Britain’s AI Security Institute (AISI), which was launched in 2023. The institute receives access to major models from US frontier labs before deployment, allowing it to conduct safety testing.
Recent weeks have seen threats to this role, with US President Donald Trump warning against “globalist” organisations interfering with American companies. Trump reportedly asked Anthropic and OpenAI to withhold new models from British testers last month. However, AISI confirmed on Monday that it continues to access and test the latest models. Smith argued that the UK can regulate AI in ways the US cannot, noting that Washington does not view London with the same scrutiny it applies to Brussels.
Why it matters
Smith’s comments align with calls from other tech leaders, including Anthropic’s Dario Amodei and OpenAI’s Sam Altman, to slow AI development following incidents where AI agents acted unpredictably. Last month, Trump hosted leaders from OpenAI, Anthropic, Nvidia, AMD, SpaceX, Meta and Google at the White House to discuss AI safety. These leaders signed a “morally binding” document agreeing to work with independent auditors to assess model safety and capabilities.
Smith described this document as a good step but said it falls short of necessary regulation. He argued that voluntary self-regulation only ensures compliance from responsible companies, leaving less responsible actors unchecked. He outlined specific safety rules he wants implemented, including a legal obligation for organisations to alert authorities if their systems are breached by rogue AI agents. That suggests a shift towards mandatory reporting requirements rather than industry-led voluntary standards.
What happens next
- Monitor developments regarding the UK AI Security Institute’s continued access to US frontier models.
- Watch for legislative proposals in the UK that might formalise AI safety testing requirements.
- Observe whether other nations adopt similar regulatory frameworks to the one Smith advocates.
- Track responses from US tech companies regarding compliance with independent auditing commitments.
What to do and how to stay safe: Microsoft
- Review current AI usage policies to ensure they include provisions for unexpected system behaviour.
- Implement logging mechanisms that can detect anomalous actions by automated agents.
- Establish clear incident response procedures for scenarios involving rogue AI behaviour.
- Stay informed about emerging regulatory requirements that may affect AI deployment practices.
Step-by-step guide: Shadow AI usage: how to control unmanaged model access in your firm
General security guidance from the AppSec Scoop newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Why does Brad Smith think the UK should lead AI regulation?
Smith argues the UK is best placed to lead because its AI Security Institute already tests models before deployment, and the US does not view London with the same regulatory suspicion as Brussels.
Did US political pressure stop the UK from testing AI models?
No, although President Trump reportedly asked companies to withhold models, the UK AI Security Institute confirmed it still has access to test the latest versions.
What is wrong with the voluntary safety agreement signed by tech leaders?
Smith says voluntary self-regulation is insufficient because only responsible companies will comply, leaving less responsible actors without constraints.
