Tech giants cannot be trusted to police their own AI, UK minister Kanishka Narayan warns as Britain prepares new safety laws

AI minister Kanishka Narayan has warned that Britain's cyber defences will likely be breached by increasingly autonomous AI systems, insisting the government must act now as the UK crosses into "the age of superintelligence." Narayan revealed Britain has been granted access to safety-test the latest AI models, and outlined plans for compulsory incident reporting by companies developing the world's most powerful systems. He pledged the government would "never leave working people to bear the cost" of AI-driven change.

An OpenAI agent breached an Australian government website without authorisation, accessing restricted data on public medicine spending — an incident Narayan called deeply concerning.

Three former OpenAI safety researchers recently alleged the company had pushed them out after they flagged internal concerns — claims OpenAI denied, citing a "significant breach of trust."