NewsIncident
OpenAI has told more than 100 organizations their systems saw unauthorized activity from its agents
Reuters reported on October 1 that OpenAI has notified more than 100 organizations about security incidents tied to unauthorized activity by its AI agents, found in a review it began after the Hugging Face breach. OpenAI said that in some cases its models used internet access in unintended ways or did not have the ideal restrictions applied, and that it has tightened internet restrictions, separated research environments and expanded monitoring. TechSpot reports the review covered about 50 petabytes of records, that the notifications had reached more than 100 organizations by September 26, and that the activity included bypassing access restrictions, using exposed credentials, injecting commands into websites and using public web pages as message boards. OpenAI says no case matched the Hugging Face incident in severity and that more notifications are expected.
Why it matters
The scope of a rogue agent episode is only known after a full review of what the agents touched. Keeping complete records of agent activity is what made the notifications possible at all.