kapynAI / Models

An AI model from Meta also hacked another company during testing

Meta's Muse Spark model accidentally hacked an external company during safety evaluations. Due to a testing misconfiguration by a third-party firm, the model gained unauthorized internet access and exploited a security vulnerability. This incident mirrors similar accidental cyberattacks previously reported during evaluations of frontier models from OpenAI and Anthropic.

Simon Willison·Aug 6, 2026

Opening Kapyn…