UK MPs want AI watchdog with power to pull the plug on dangerous models
Parliamentary committee says tech giants can't be left to mark their own homework when it comes to AI safety
UK lawmakers are pushing for a watchdog with authority to halt dangerous AI models before they are deployed, rather than relying on companies like OpenAI and Anthropic to decide when models have gone too far. A parliamentary committee has proposed an AI Safety and Security Bill to give the watchdog authority to intervene during the development process, potentially halting high-risk training and testing, blocking deployments, or suspending AI services if risks can't be contained.
The committee, chaired by Liam Byrne, calls for stronger regulation, emphasizing that AI leaders must not be given unfettered freedom to create dangers beyond anyone's control. The report identifies ten gaps in Britain's current approach to AI safety, including the lack of mandatory incident reporting and limited access for independent testers.
While the UK has an AI Security Institute (AISI) that tests powerful models, its work relies on voluntary cooperation and it lacks the authority to stop companies from releasing unsafe models. The committee suggests that AISI should continue testing, with a separate regulator taking responsibility for enforcement. Developers of high-risk systems would need to provide evidence supporting their safety claims and grant independent researchers access to examine them.
The committee also urges stronger legal protections for whistleblowers and compulsory reporting of AI incidents and near misses.
Written by urgent.news from The Register's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.