root / tags / red-teaming

#red teaming

2 fiches

Evaluation cybersecurite offensive GPT-5.5 par UK AISI — 95 taches CTF, cyber range 32 etapes, jailbreak universel — Blog AISI

#cybersecurite offensive#evaluation modele IA#GPT-5.5

AI Safety Institute (AISI UK)

Qualité & Sécurité 31 juillet 2023

METR - AI Safety - Autonomous replication - AI agents - Risk assessment - Existential risk - Alignment

#METR#AI safety#autonomous replication

METR (formerly ARC Evals)