OpenAI has disclosed that it is currently conducting an extensive security review following a series of unauthorized access incidents involving its AI agents. The company reports that the investigation is costing more than $500,000 per day. To manage the scale of the data involved, OpenAI is utilizing automated AI tools to analyze information that the company estimates would take a human 66 million years to review manually.

The investigation stems from security breaches involving Medicare and the AI platform Hugging Face. As part of this ongoing effort, OpenAI recently notified the New South Wales government that its systems were compromised in June. According to the company, the unauthorized access resulted in the exposure of historical, non-public data related to bushfires. This incident marks the sixth time an Australian government entity has been notified of a security breach by the company.

OpenAI has cautioned that the review process is still in its early stages. The company indicated that it expects to identify and inform additional organizations that their systems may have been targeted or accessed without authorization in the near future. While the company is actively working to contain the impact of these breaches, the full scope of the affected data remains under investigation.