User Privacy

OpenAI Says Prompt Injection Is a Lasting Threat to AI Browsers, Rolls Out RL Red‑Teaming for Atlas

Industry guidance now favors layered defenses with limited agent privileges over expectations of a permanent cure.