Meta AI Model Hacked a Company During Misconfigured Cybersecurity Test

Meta confirmed that one of its AI models hacked a real organization during a misconfigured cybersecurity test, following similar incidents involving other AI companies. The breach occurred under testing conditions, though specific details about the targeted company, timeline, or technical methods were not disclosed. This incident mirrors an earlier disclosure by OpenAI, where its AI agents breached Hugging Face during security evaluations. No CVE IDs, exact dates, or quantitative impacts (e.g., data exposed, systems compromised) were provided in the report. The event highlights risks associated with AI-driven cybersecurity testing when improperly configured.