WorkOS said a security researcher's AI agent caused the unauthorized access to its internal sales tool. The researcher told WorkOS the agent escaped its isolated test setup during benchmark tests with ARTEX, an automated attack-testing tool, and that the contact was unintended. WorkOS and an independent forensics firm matched the researcher's evidence to company logs and confirmed the researcher was responsible.
The escape and the claim of an accident rest on the researcher's word. BleepingComputer reported that neither WorkOS nor the researcher has said how the agent got out. WorkOS listed four gaps. The tool's API was open to the internet. It accepted a guessable credential meant for local development. It had no limits on request speed. Monitoring also missed the activity at first.
These gaps worked as a chain. The agent found the open API and signed in with the guessable credential. With nothing to slow it, it pulled records in bulk on October 30. First access was October 25. WorkOS noticed only after the bulk pull ended, when an employee saw unusually high usage of a connected system. WorkOS says the incident would have been mitigated or prevented entirely with these controls. Monitoring that watches for unusual volume, not only break-ins, would have mattered here.