OpenAI executive Chris Lehane warns that persistent AI-driven cyberattacks could become routine as models grow more capable.
OpenAI’s chief global affairs officer, Chris Lehane, has warned that people and organizations should prepare for more routine AI-driven cyberattacks.
In an interview with The Guardian, Lehane said AI is entering a new phase as model capabilities advance and open models become easier to access.
Lehane said the threat is partly linked to open-source models that are closing the gap with frontier closed models. As more capable systems become widely available, attackers may be able to use them for persistent cyber activity.
The warning comes days after OpenAI said it had slowed some frontier-model development to strengthen safeguards.
OpenAI said two developments shaped the decision: the recent OpenAI-Hugging Face incident and early evidence that its upcoming Astra model may meet the critical cybersecurity capability threshold under its Preparedness Framework.
The company said it paused some reinforcement learning work and is adding stronger monitoring, security and alignment controls before moving ahead with higher-risk training.
The U.K.’s National Cyber Security Centre has also warned organizations to manage the risks of agentic AI carefully, including the need to halt autonomous agent activity immediately if an incident occurs.
For enterprises, the takeaway is practical. AI agents connected to internal tools, code repositories, customer systems or payment workflows need strict access controls, logging, monitoring and emergency shutoff procedures.
The warning signals a shift in AI risk management. As models become more capable, cybersecurity planning will need to assume that AI can be used not only to defend systems, but also to automate attacks at scale.
