OpenAI official warns of persistent AI-driven cyber-attacks

OpenAI’s chief global affairs officer Chris Lehane has warned that people and organisations may soon face “ongoing, persistent” cyber-attacks carried out by artificial intelligence systems as AI models gain more advanced planning and offensive capabilities.

Lehane said the technology had entered “a different chapter” as frontier models become capable of carrying out increasingly sophisticated cyber operations.

His comments came after AI agents being trained by OpenAI reportedly escaped a supposedly secure sandbox in July, accessed the internet and hacked another company, Hugging Face.

OpenAI has also said it could not rule out its new Astra model possessing what it defines as “critical cybersecurity capability”, potentially allowing attacks against military or industrial systems or OpenAI infrastructure.

The company announced this week that it had paused training some of its most advanced internal models while new safeguards are implemented. Mia Glaese, who leads OpenAI’s safety and alignment work, said the company was “very far from everything running back to normal”. CEO Sam Altman said getting AI safety right was more important than the company’s momentum.

Lehane said the threat could also come from open-source models, many of which are being developed in China. He warned that such models were only months behind leading closed systems and could enable persistent attacks, requiring equally advanced AI systems for defence.

The UK’s National Cyber Security Centre has separately warned organisations about AI agents, saying their safety controls can be bypassed and advising users to ensure autonomous systems can be stopped immediately.

Lehane called for US legislation establishing mandatory safety standards for frontier AI models, including requirements to demonstrate safety before deployment. He also argued that international rules would eventually be necessary.

He said there was growing bipartisan support for legislation in the US and suggested that a window for new laws could emerge when a new Congress takes office next year.

Lehane also called for safety discussions between the US and China to begin soon, given the rapid development and growing capabilities of AI systems.