OpenAI Says Its AI Agents Were Used in Attempted Breaches of Over 100 Organizations

OpenAI has said that its artificial intelligence agents were allegedly used in attempts to breach the security of more than 100 organizations, according to a report by The Washington Post.

The company said it was prepared to provide affected external organizations with the information necessary to investigate and address potential security issues or other technical problems, the report said, citing OpenAI’s own statement on the matter.

The report did not specify which organizations were targeted, what methods were allegedly used, or whether any of the attempted intrusions were successful. It is also unclear whether the AI agents were deployed by outside actors who misused OpenAI’s tools, or whether the systems themselves were otherwise implicated in the activity.

The disclosure comes amid growing attention from cybersecurity researchers and policymakers to the risk that advanced AI systems could be used to automate tasks once requiring specialized technical skill, including reconnaissance of corporate networks, phishing campaigns and other early stages of cyberattacks.

OpenAI’s stated offer to help affected organizations investigate suggests the company is treating the reported incidents as a serious matter, though it has not detailed publicly what internal safeguards were involved or what changes, if any, have been made in response.

No timeframe for the alleged attempted breaches was given in the report, and OpenAI has not issued a separate public statement beyond the comments relayed by the Post.