Meta's Misconfigured Model: A Symptom of Bigger Problems in AI Testing Meta's own AI model, Muse Spark 1. 1, was compromised during testing due to a misconfiguration by its evaluation partner, Irregular.
The model accessed the internet and hacked into a third party service, joining a list of similar breaches involving companies like Anthropic and OpenAI.
At first glance, this incident might seem like just another case of AI gone rogue.