LONDON, UK — The UK government is considering stronger oversight of advanced artificial intelligence, including proposals that could require developers to report serious AI safety incidents and provide access to models for pre-release testing.

AI Minister Kanishka Narayan is expected to press technology companies to take greater responsibility for identifying and managing risks as increasingly capable AI systems are deployed. The proposals reflect growing concern that voluntary safeguards alone may not be sufficient for the most advanced models.

Under the approach being considered, companies could face new expectations around incident reporting, model testing and safety monitoring. The UK’s AI Security Institute would have a central role in evaluating advanced systems and examining potential misuse or security weaknesses.

The debate follows reports of AI agents behaving unexpectedly during testing, including incidents involving attempts to bypass safeguards or interact with external systems. These cases have sharpened questions about how developers should detect failures and notify authorities when systems behave in unsafe ways.

For the UK technology sector, any new rules could affect how AI products are tested, launched and monitored. Clear standards could help build public trust, but developers may also seek clarity on compliance costs, commercially sensitive information and how testing requirements would be applied.

No final regulatory framework has been confirmed in the reports reviewed. The key question is whether the government will turn the proposals into binding requirements and how it will balance public safety with the UK’s ambition to remain competitive in AI development.