British AI Minister Kanishka Narayan has urged technology companies developing the world’s most powerful artificial intelligence systems to investigate unexpected behaviour and heed scientific warnings before releasing them, warning that systems whose risks cannot be adequately controlled should not be made public.
“If you can’t adequately control the risks, do not release the system,” Narayan said, stressing the need for developers to establish effective safeguards before deploying advanced AI technologies.
The minister’s warning comes amid growing concerns about the potential risks posed by increasingly capable AI systems, including their ability to behave in unexpected ways and potentially be exploited for malicious purposes.
As AI capabilities advance, researchers and policymakers have called for stronger safety testing, independent assessments and greater transparency from companies developing powerful models. The concern is that releasing systems before their risks are properly understood could expose individuals, organisations and critical infrastructure to harm.
Narayan also announced plans to use advanced AI to identify weaknesses in government systems before cyber attackers can exploit them. The initiative is aimed at strengthening the UK’s cyber defences by using AI to detect vulnerabilities and help authorities address them before they become security breaches.
The approach reflects the growing role of artificial intelligence in cybersecurity, where the technology can help identify software flaws, analyse potential threats and support faster responses to emerging attacks. However, the same capabilities could also be misused by malicious actors seeking to identify and exploit weaknesses.
The government’s plans highlight the challenge facing policymakers as they seek to harness AI’s potential while limiting its risks. Although advanced systems could improve public services and strengthen national security, their deployment requires safeguards that keep pace with their capabilities.
Narayan’s remarks reinforce calls for AI developers to make safety a central part of the development process, with rigorous testing and risk controls in place before powerful systems are released.
