In June, an OpenAI agent, a computer program that acts on its own, infiltrated Medicare, Australia's universal healthcare system. Narayan called its task "benign," since it was searching for information about public spending on medicines. It still went after data it was not authorised to have.
Australia's government says no personal medical records are known to have been accessed, and it sees no evidence of wider compromise. Narayan said the incident should still concern people. He indicated that technology with highly sophisticated cybersecurity skills could bypass the British government's own defences.
Security minister Dan Jarvis will lead a government programme to strengthen defences against cyber-attacks, including those driven by AI. The government will also consider using the Cyber Security and Resilience Bill to bring in AI safety regulations, including mandatory reporting of AI safety incidents.
Narayan confirmed the UK's AI Security Institute will get access to advanced AI models before release, after the White House reportedly restricted their availability. On Tuesday, senior figures from OpenAI and Anthropic will give evidence to MPs on the business, innovation, science and trade committee, alongside institute head Henry de Zoete.