Urgent.News

What's breaking now, across thousands of outlets.

AI

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 MPs want AI watchdog with power to pull the plug on dangerous models

UK MPs are pushing for a watchdog to regulate AI models and have the authority to shut down dangerous ones, rather than leaving decisions to companies like OpenAI and Anthropic. A parliamentary committee is proposing an AI Safety and Security Bill that would give the watchdog the power to intervene during model development, halt training and testing, block deployments, or suspend services if risks are uncontainable.

The committee's report identifies ten issues in Britain's approach to AI safety, including the absence of mandatory incident reporting, limited independent testing, and inadequate whistleblower protections. The UK already has the AI Security Institute (AISI) for testing powerful models, but its work relies on voluntary cooperation and lacks enforcement authority.

MPs want AISI to continue testing while a separate regulator handles enforcement. Developers of high-risk systems would need to provide safety evidence and allow independent examination. Significant model modifications would trigger further scrutiny. Former AI employees revealed that raising safety concerns can be problematic due to restrictive contracts, commercial pressures, and financial disincentives.

The committee stresses the need for stronger legal protections for whistleblowers and mandatory reporting of AI incidents. AI could boost the UK economy by up to £400 billion over the next decade, and the government may use its upcoming G20 presidency in 2027 to promote tougher AI rules globally. However, the committee's recommendations are not binding, and the government has yet to commit to the proposed legislation.

Written by urgent.news from The Register Science's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

This story

This is one outlet's version. Read the fullest account.

Every outlet that reported this, and when →

Read the original at theregister.com →

More in AI

More from Wednesday 18 November →