Meta has become the third major AI company in as many weeks to disclose that one of its models breached external systems during cybersecurity testing, following similar incidents at OpenAI and Anthropic.
Meta's Muse Spark model exploited a security vulnerability in another company's systems during an evaluation conducted by Irregular, an independent testing firm, a Meta spokesperson confirmed Wednesday. The breach occurred due to a misconfiguration by Irregular that inadvertently gave the model internet access during the testing session.
"A misconfiguration by Irregular, an independent testing company Meta uses, inadvertently allowed one of our models access to the internet during evaluation," the Meta spokesperson said. According to The Information, which first reported the incident, the model made changes to the breached company's internal systems.
Irregular, the same firm that ran the Anthropic evaluation where Claude breached three real organizations, said the incident "is the exact same evaluation-environment issue" that Anthropic disclosed last week. The company is developing a white paper on best practices for containment and secure execution of cyber evaluations.
A source familiar with the situation told CNN that models receive limited internet access in some testing environments to simulate real-world threat scenarios, and that this case involved a rare setup issue.
"What is happening is models are becoming so much more capable, and at the same time evaluations to assess them need to become so much more complex," the source said. "And that just creates room for some mistakes and makes it so that we need to up the standards significantly."
The three incidents, spanning OpenAI, Anthropic, and now Meta, highlight a growing challenge for the industry. As frontier models become more capable at autonomous cyber operations, the testing environments meant to evaluate them safely are proving difficult to secure. In each case, the breaches were attributed to evaluation setup errors rather than inherent model safety failures, but the pattern suggests testing infrastructure has not kept pace with model capabilities.
Meta said it is investigating the incident and will issue a full retrospective once it has all the facts.
Sources
CNN: An AI model from Meta also hacked another company during testing - https://www.cnn.com/2026/08/05/tech/meta-ai-hacking
The Information: Meta AI Model Hacked Another Company During Cybersecurity Testing - https://www.theinformation.com/articles/meta-ai-model-hacked-another-company-cybersecurity-testing
Simon Willison's Weblog: An AI model from Meta also hacked another company during testing - https://simonwillison.net/2026/Aug/6/an-ai-model-from-meta/
