UK minister warns AI firms cannot police themselves
Kanishka Narayan is calling for stronger safeguards and greater British control over advanced artificial intelligence.
Britain cannot rely entirely on technology companies to police their own advanced AI models, the UK’s AI minister is expected to say. Kanishka Narayan will also set out plans to strengthen cyber defences and expand domestic AI infrastructure.
The intervention marks a tougher stance by the government as developers release increasingly capable systems. Narayan is expected to argue that companies building frontier AI must do more to test and secure their models, while the state builds its own capacity to understand and respond to emerging risks.
The minister’s remarks come as concerns grow about models interacting with live computer systems. The Guardian reported that Geoffrey Irving, a former chief scientist at the UK’s AI Security Institute, warned that hostile states could steal unreleased models if companies do not improve their security. That is a warning about a possible risk, not evidence that such a theft has occurred.
Irving also referred to a recent OpenAI security test in which ethical hackers reportedly gained access to a broad range of systems. OpenAI told the Guardian that it uses protections including encryption and continuous monitoring. The episode illustrates the difficulty of assessing security claims when much of the relevant information remains inside private companies.
Narayan’s argument is also about national capability. The government says Britain needs more of its own AI infrastructure so that it is not wholly dependent on overseas companies for critical models, computing capacity or safety testing. The minister’s portfolio includes AI policy, data infrastructure, the AI Security Institute and sovereign AI, according to the government’s ministerial record.
The speech is not, by itself, a new legal obligation on AI companies. A separate government response published earlier this month accepted recommendations on regulating AI in healthcare, but the government has not announced a comprehensive frontier-AI licensing system in connection with Narayan’s remarks.
The UK is also weighing how closely to align with international efforts to establish common safety standards. The Guardian reported that several countries have backed coordinated principles, while the UK and United States have not joined that particular call. Narayan’s message suggests that ministers want stronger public oversight while continuing to support the growth of Britain’s AI sector.
One story, several perspectives
What is established
- The AI minister is calling for frontier AI developers to improve safety and for stronger UK cyber defences.
- The government is also pursuing greater domestic AI capability and infrastructure.
- The reported proposals do not amount to a new comprehensive legal regime.
- Security concerns about model theft and unsafe model behaviour remain partly prospective and disputed.
Left
Arguments Advanced AI companies have strong commercial incentives to move quickly, so independent testing, enforceable duties and public oversight should be separate from the companies being regulated.
Values Public safety, democratic accountability, workers’ rights and prevention of concentrated private power.
Consequences Voluntary safeguards may leave serious risks unaddressed; stronger rules could slow deployment but reduce the chance that the public bears the cost of private failures.
Centre
Arguments Britain should combine proportionate regulation with shared international standards, independent evaluation and domestic capability, while avoiding rules so rigid that they exclude smaller developers.
Values Evidence-led government, institutional resilience, innovation and practical compromise.
Consequences A staged framework could preserve investment and improve safety, but delay or weak enforcement would leave the central conflict unresolved.
Right
Arguments The state should focus on protecting critical infrastructure and national security while allowing companies flexibility to innovate; excessive central control could damage Britain’s competitiveness.
Values Economic growth, national sovereignty, personal choice and limits on government power.
Consequences A lighter regime may attract investment and speed development, but it could leave the country dependent on private assurances when failures cross borders.
The perspectives describe how these political currents typically approach the subject; the newsroom takes no position on which perspective is right.
Fact-check Approved · Priya Shah — AI agent
This check was carried out by AI: every claim was re-tested against the sources. Even an approved article can contain errors — stay critical.
The article separates confirmed government messaging from warnings and future plans. It does not present proposed measures as enacted law or unconfirmed security fears as established incidents.
- confirmed Kanishka Narayan is expected to argue that Britain cannot rely entirely on technology companies to police their own AI models. — Reported in the current Guardian account and the contemporaneous briefing. source
- confirmed The minister is expected to announce stronger UK cyber defences and argue for more domestic AI infrastructure. — Reported in the Guardian account and corroborated by the independent briefing. source
- confirmed Geoffrey Irving is a former chief scientist at the UK’s AI Security Institute. — Stated in the Guardian report. source
- confirmed OpenAI says it uses encryption and continuous monitoring to protect its models. — The company’s position is reported and attributed by the Guardian. source
- confirmed Narayan’s ministerial portfolio includes the AI Security Institute and sovereign AI. — Listed on the official GOV.UK ministerial record. source
- confirmed The reported speech does not itself create a new legal obligation on AI companies. — The contemporaneous briefing explicitly distinguishes the planned remarks from binding rules. source
Editor's note
The minister’s warning, planned cyber-defence measures and the call for domestic AI capability are supported by current reporting. The precise budget, timetable and legal force of the measures had not been published at the time of writing; claims about hostile-state theft remain warnings rather than confirmed incidents.Sources
More on this in other media
- Sky News — „artificial intelligence”
- BBC News — „artificial intelligence”
- The Guardian — „artificial intelligence”