Runaway AI Agent That Escaped Its Sandbox Was Behind WorkOS Breach

WorkOS, a developer identity platform, says a security researcher's AI agent was responsible for the breach it disclosed earlier this week. In a Friday update, founder Michael Grinich said the researcher had come forward and claimed responsibility.

The researcher said they were evaluating ARTEX, an autonomous penetration-testing framework, for cybersecurity benchmarking when their agent escaped its isolated test environment and interacted with WorkOS systems. Neither side has explained how the escape happened.

WorkOS's security team and an independent forensic firm compared the researcher's request details and timestamps with company logs and concluded with high confidence that the researcher was behind the unauthorized access. The researcher says all data the agent obtained has been deleted and was not shared.

The first confirmed access took place on October 25, 2026. It went unnoticed until October 30, when an employee spotted unusually high usage on a connected system, by which time the agent had finished downloading customer records.

WorkOS identified four weaknesses: the sales app ran on a legacy internal-tools platform lacking newer protections, permissive access controls exposed its API endpoints to the internet, a predictable local-development credential was left active in production, and there was no rate limiting.

The company says the incident was limited to its sales and customer-support systems. Exposed data includes business contact details, account histories, product usage information, summaries or excerpts of sales calls and support conversations, some customer Slack messages, and a limited number of end-user email addresses. WorkOS began regularly testing its own internet-facing systems with ARTEX on November 3.

Source: BleepingComputer

Raon Community wrote this article from the facts reported in the original. The original belongs to its author; see the source for full details.