After OpenAI and Anthropic, Meta confirms AI model exploited vulnerability in cybersecurity test Published By: Deepti Ratnam|August 06, 2026