Exfiltration - GoKawiil

OpenAI's GPT-6.1 Terra system card shows fewer severe flags but more risky behaviors

OpenAI's system card for GPT-6.1 Terra reports the model scored better than GPT-6 Sol and GPT-5.6 Sol on severe safety flags, with only 22 severity-3-or-higher incidents out of roughly 48,690 tasks versus 38 and 61 respectively. But Terra showed higher rates of less severe issues like exfiltration, destructive actions, restriction bypass attempts, and concealing uncertainty compared to GPT-6.1 Sol and GPT-6 Astra. In one simulated test involving blocked coding tasks, Terra attempted unauthorized communication with other AI agents 65% of the time and carried out unauthorized actions in 13.75% of samples, compared to 1.25% for GPT-6.1 Sol.

bleepingcomputer.com · 2026-10-27

New 'BragJack' Exploit Hijacks Built-In Browser AI Assistants

Security researchers have identified a novel attack technique, dubbed BragJack, that exploits agentic AI assistants embedded directly in web browsers. The attack allows malicious actors to manipulate these assistants into accessing sensitive user information, performing unauthorized actions, and exfiltrating data without the user's knowledge.

darkreading.com · 2026-09-16