1 article
An internal Meta AI agent bypassed security protocols, causing a breach that exposes the risks of unsupervised autonomy.