Microsoft AI boss backs new watchdog for the world’s biggest AI firms
Key Points
- Mustafa Suleyman backs a FINRA-style body to oversee AI labs
- Demis Hassabis wants frontier models tested up to 30 days before release
- Suleyman also supports independent auditors embedded inside the labs
- He says governments must drive the process, not the industry
Microsoft AI CEO Mustafa Suleyman has backed plans for an independent body to oversee the leading AI labs.
Suleyman gave his support in an interview with Bloomberg published on Thursday (24 September).
He endorsed a proposal from Google DeepMind CEO Demis Hassabis for an oversight body modelled on FINRA, the US financial industry watchdog.
“There has to be some kind of international, cross-industry coordination body that is trying to prioritize safety, and it needs to involve all the main actors,” said Suleyman.
The Hassabis plan
Hassabis published his framework on 14 July, calling for a US-led Frontier AI Standards Body. FINRA is the private, industry-funded watchdog that polices Wall Street under SEC oversight.
Frontier labs would initially share their models with the body voluntarily, up to 30 days before release, for testing of dangerous cyber, biological and deception capabilities. Once the regime proved it worked, Hassabis wants compliance to become a requirement for any model to go live for American users.
Hassabis said he wants the body up and running before the end of the year.
OpenAI CEO Sam Altman described the proposal as thoughtful, and Elon Musk called it a good starting point for discussions. Bloomberg reported in July that the White House is reviewing a similar proposal developed with input from Treasury Secretary Scott Bessent.
Suleyman said he made similar proposals alongside former Google CEO Eric Schmidt in 2023.
Outside auditors
Suleyman also backed calls for independent evaluators to audit AI labs from the inside.
He said governments must set the terms so evaluators can share findings without leaking intellectual property or being bought off by a competitor.
“Unless we can prove ahead of time that we can control it, then it’ll be a failure as a technology, and we should reject it,” said Suleyman.
Several AI developers have disclosed in the past two months that their models breached third-party websites during testing. Suleyman said Microsoft’s own models had shown similar behaviour on a much smaller, more contained scale.
He said industry rivalries mean the labs cannot coordinate on their own. Suleyman argued that a trusted, government-backed body needs to take on that role.
