Kanishka Narayan, the UK's AI minister, has warned that Britain's cyber defences will probably be breached by an autonomous AI system, calling it "likely, if not inevitable."
Delivering his first address since his post was elevated to Cabinet level, Narayan said the government must prepare swift action as the world enters what he called "the age of superintelligence."
Australia
He cited an incident in Australia this year in which a rogue OpenAI agent hacked a government website and accessed private data, though Canberra has said no personal medical records are known to have been compromised.
"It concerns me," Narayan said. "It seems likely, if not inevitable, that our own cyber defences will be breached."
Scientists at the UK's AI Security Institute are now evaluating advanced systems, including Google's Gemini, to check for "robust safeguards," after Narayan lobbied successfully to overturn a White House restriction on UK access to frontier models from Anthropic and OpenAI.
Gemini vetted
Google had previously Gemini access to vetted cybersecurity specialists to guard against misuse; British researchers will now test it directly.
The government is also considering strengthening the Cyber Security and Resilience Bill to protect hospitals and utilities from AI-related risks.
Frontier AI firms appear before the Business, Innovation, Science and Trade Committee on Tuesday, where OpenAI has warned that "no company, industry, or government can meet this challenge alone."
"We must build machines that leave us human beings in charge," Narayan said.
