Microsoft President Urges AI Safety Rules, Sees Uk As Leader
Brad Smith says voluntary commitments are insufficient and argues Britain is well placed to lead AI safety oversight as the US favors industry-led measures.
Microsoft president Brad Smith is calling for stronger rules governing artificial intelligence, arguing that AI systems face less safety oversight than everyday consumer products despite the risks they pose. According to TechCrunch and Engadget, he sees the UK as particularly well placed to lead on AI regulation while the US favors an industry-led approach.
Table of Contents
Why Smith wants binding rules
Smith compared AI oversight with the inspections and safety requirements applied to milk and washing machines. TechCrunch reported that, on a BBC podcast, he described AI as both a tool and a potential weapon that could put global cybersecurity at risk. Engadget reported similar remarks from an ITV interview, in which Smith said the time had come to address risks from powerful models.
Smith’s concern is not only whether major developers will act responsibly, but whether every company will. According to TechCrunch, he welcomed a recent voluntary agreement by technology leaders as a step forward, while arguing that self-regulation cannot ensure that less responsible companies follow the same practices.
What the proposed safeguards would cover
TechCrunch reported that Smith wants organizations to be legally required to notify others if rogue AI agents breach their systems. He also spoke favorably about outside evaluation before AI products are released and about mechanisms to pause or slow an AI system if people lose control of it. These were proposals Smith discussed, not rules the reports say have been adopted.
The debate follows a White House meeting at which technology leaders agreed to work with independent auditors to assess their models, according to TechCrunch. Engadget described the pact as emphasizing self-regulation and safeguards for advanced models. It also reported criticism that the arrangement leaves companies too much power to decide what counts as safety.
Why the UK features in the debate
Smith argued that London has a political opening Washington might not offer Brussels, according to both outlets. TechCrunch pointed to Britain’s AI Security Institute, launched in 2023, which has received access to major US-developed models for testing before deployment. The outlet reported that President Donald Trump had reportedly asked Anthropic and OpenAI last month to withhold new models from British testers. It also said the institute confirmed on Monday that it could access and test the latest models.
For US AI developers, the practical question raised by Smith’s proposals is whether safety testing and incident reporting should remain voluntary or become legal obligations. Neither report describes a new binding rule. They do, however, show a prominent technology executive urging governments to set requirements that would extend beyond the companies already willing to make safety commitments.
Sources
This story was compiled by AI from the reports below. Read the originals for the full details.