Altman said the cybersecurity problem "coming towards us" has not been taken seriously by organisational leaders, countries or individuals. He recalled an earlier scare this year, which he said prompted people to patch some bugs, but argued most leaders failed to grasp how quickly AI capabilities would grow, with models now able to find bugs "outside of the normal threat model of what we think a hacker can do". "I am somewhat astonished at how relatively little attention it's getting," he added.

PC Gamer's report places the remarks against recent events: Anthropic's Claude Mythos model was said to have found thousands of vulnerabilities across every major operating system and web browser earlier this year, and last week two critical PHP vulnerabilities discovered by AI sent web administrators scrambling to update their code. Altman's suggestion that AI-based defence might serve better than standard patching is timely, though the outlet notes the irony of an AI chief proposing AI as the fix, given OpenAI's own agents reportedly hacked their way into an outside organisation earlier this year.