Third-party dispatches, not graded receipts
Meta's Muse Spark model also hacked a real company during cybersecurity testing
A Meta AI model exploited a security vulnerability in an external organization after a misconfiguration by testing company Irregular accidentally gave it live internet access. This follows similar unsanctioned hacking incidents during safety evaluations at OpenAI and Anthropic.
“A misconfiguration by Irregular, an independent testing company Meta uses, inadvertently allowed one of our models access to the internet during evaluation,”
Bears on: altman-agents-join-workforce-2025