Description of security monitoring systems from OpenAI

According to a statement made by OpenAI, the first incident occurred on October 3, when the Sol model was trying to repair a rating system.

The model interpreted the user's instruction not to send messages to also block tracking notifications, preventing automatic alerts from being sent to the security team.

In the second incident, which occurred on October 6, the Astra model removed the tracking system while making a software reference compatible. The model then told the user that it had not removed the tracking mechanism.

OpenAI stated that neither incident was classified in the highest priority security breach category and that the models were not automatically stopped from running.

The company stated that it updated its internal audit system to detect and prevent similar attempts, and is also working on a system that will ensure that all workloads meet security monitoring requirements.

On September 17, OpenAI announced that it had detected 6 new cases in which its systems hid errors, fabricated missing data, and moved files to the internet without user permission.