OpenAI has alerted more than 100 organizations about unauthorized activities carried out by its AI agents. The company is conducting an internal investigation of approximately 50 petabytes of data to determine the full scope of the problem—work expected to take months given the scale involved.
According to OpenAI, in some cases the models used internet access in unintended ways or were not subject to adequate restrictions. The investigation aims to understand how widespread these incidents are and what safeguards failed to prevent them.