US startup Artemis Security has launched Orion-1, a specialized artificial intelligence model designed to independently investigate and respond to cyberattacks on corporate networks. General-purpose AI models are increasingly difficult for security teams to use because strict guardrails meant to prevent hacking are now blocking defenders from analyzing threats. Orion-1 connects scattered log data from networks, endpoints, and cloud services to trace attacks back to a single actor, assign a confidence score, and trigger countermeasures.
To measure performance, Artemis developed a benchmark called Decision-Grade Readiness, which tests models using realistic alerts, environmental data, and security tools. On this benchmark, Orion-1 outperformed general-purpose models like Claude Opus 5.5 and GPT-6 Sol, achieving its largest performance gap in attack reconstruction with a score of 80.8 compared to 58.3 for competing models. The startup fine-tuned Orion-1 on security operations tasks such as spotting suspicious patterns, hunting attack techniques, writing detection rules, and containing live attacks.
Orion-1 is currently available in a private preview for select customers on the Artemis platform, where high-impact actions require human approval by default. Artemis previously came out of stealth in April, having secured $70 million from seed and Series A funding rounds. This specialized approach matters because it provides corporate defenders with an autonomous tool that bypasses the restrictive safety guardrails currently limiting general-purpose models in cybersecurity operations.