\n\n

Key developments:

• Meta rolled out an internal "Red‑Team AI Agent" that continuously attacks its own defenses, aiming to surface blind spots before malicious actors can exploit them. This autonomous

…

\n\n

Key developments this week show AI red‑team­ing moving from niche testing into systematic, organization‑wide practice. A Microsoft Research study demonstrated that focused jailbreak probes can uncover model weaknesses with …