📰 Key Takeaways

Microsoft President Brad Smith publicly called for AI regulation to happen now in a recent BBC podcast, pointing out that even consumer products like milk and washing machines face more oversight than the companies building AI. He described AI as both a “weapon and a tool” that’s putting global cybersecurity at risk, saying “the time has come” to establish global safety rules.

Smith specifically named the UK, rather than the US or EU, as best positioned to lead on global AI safety rules. The UK’s AI Safety Institute (AISI), established in 2023, has quickly become a leader in AI safety testing — it gets access to test frontier lab models from the US before they’re formally deployed. But that role has recently come under threat: US President Trump has warned about “globalist” organizations interfering with American companies, and reportedly asked Anthropic and OpenAI last month to pause giving UK testers access to new models, though AISI confirmed on Monday that it currently still has access to test the latest models. Smith believes the UK can do regulation that the US can’t precisely because Washington doesn’t view London with the same political hostility it reserves for Brussels — and that’s the UK’s advantage.

These remarks come as tech leaders like Anthropic’s Dario Amodei and OpenAI’s Sam Altman have called for slowing down AI development, against a backdrop of multiple AI agent incidents going off the rails. Last month, Trump convened leaders from OpenAI, Anthropic, Nvidia, AMD, SpaceX, Meta, Google and others at the White House to discuss AI safety, where all parties signed an “ethically binding” document agreeing to work with “independent review bodies” to assess the safety and capabilities of AI models. Smith called this a “good first step,” but said there’s still a gap before we get the regulation that’s actually necessary — the problem with voluntary self-regulation is that only the “good companies” follow it, while the companies you really need to worry about are the less responsible ones. He laid out several concrete proposals, including a legal requirement to report when systems are breached by AI agents, external review bodies approving AI products before they launch, and support for a “kill switch” to pause or slow down AI systems if humans lose control.

💬 JudyAI Lab Take

Microsoft’s Smith is publicly demanding AI regulation, and he’s pointing at an uncomfortable reality — even milk and washing machines get more scrutiny than AI companies. This isn’t fearmongering; it’s the real gap at the world’s frontier labs.

For AI builders, what’s worth holding onto here isn’t “regulation is coming” — it’s the judgment behind it: voluntary self-regulation only constrains the “good companies,” and does nothing for the ones you actually need to worry about. Smith’s proposals — mandatory breach reporting, external review before launch, a “kill switch” — all point to the same design philosophy: safety shouldn’t be a patch applied after the fact, it should be a built-in accountability mechanism from the architecture stage onward. Whether you’re building an agent, a tool, or a product, the questions “can this be inspected externally?” and “can this be paused?” will eventually need answers.

Next time you’re designing an AI system, ask yourself: if it had to face outside scrutiny today, is your system ready to be looked at?

📅 Source Info

🔗 Further Reading