US startup Artemis Security has put Orion-1 into a private preview for a handful of customers. The model reads logs from authentication services, cloud setups, endpoints, SaaS apps and network gear, links related signals to a single attacker, and then gives a confidence score and response recommendation. Artemis says Orion-1 beat Claude Opus 5.5 and GPT‑6 Sol on its own “Decision‑Grade Readiness” benchmark, scoring 80.8 points in attack reconstruction versus 58.3 for the runner‑up.
Why it matters
Security teams get a model that can auto‑correlate attack data and suggest actions, reducing reliance on over‑restricted general AI tools.
