Eigen RadarAI
Analysis

UK lawmakers propose powers to halt high-risk AI training

A committee of UK lawmakers wants a regulator to be able to intervene while companies are still building high-risk artificial intelligence models. Its proposal would also give authorities power to stop deployment or suspend a service when developers cannot manage the risks. MPs are asking for a safety bill in the next parliamentary session.

Artificial Intelligence··Night
In an AI training data hall in London, a powered-down server rack is locked out with a red padlock while the racks behind it keep running.

A regulator could step in before release

In the UK, a parliamentary committee on business, innovation, science and trade wants the government to establish a regulator that can intervene during high-risk artificial intelligence training and testing. The lawmakers propose allowing it to block deployment or suspend services when companies cannot control the risks. They also call for a fresh review if a model’s capabilities or safeguards change substantially.[1], [2]

Model testing relies on voluntary access

The committee heard from researchers, former employees and representatives of four major developers, including OpenAI and Google DeepMind. The companies said they would hold back models they consider unsafe. MPs pointed out that companies apply different standards. The UK’s AI Security Institute can test models when companies cooperate, but it has no power to stop a release.[1]

MPs want companies to report near misses

Former employees described obstacles to speaking up about safety, including contractual restrictions and commercial pressure. The committee recommends legal protections for people who raise concerns and mandatory reports of both incidents and near misses. It wants an AI Safety and Security Bill in the next parliamentary session. MPs also proposed using Britain’s G20 presidency in 2027 to seek common international rules on testing, incident disclosures and high-risk releases.[1]

References

  1. News sourceThe RegisterUK lawmakers propose a regulator with powers to intervene in AI development↩1↩2↩3
  2. News sourceOn The WireUK committee calls for an AI safety law and a statutory regulator↩