Arcjet AI security models

Arcjet Himalia 1 is a custom security model family built to analyze agent activity and detect prompt injection attacks. Key capabilities: • General purpose prompt injection detection, live for every SDK call that uses prompt injection detection • Arcjet Himalia 1 as the default model for prompt injection analysis • Arcjet Thebe 1, a specialist model for coding agents • Arcjet Carme 1, a specialist model for autonomous security agents • Custom fine-tuned models built for threats that general models handle poorly. Arcjet's prompt injection detection now runs on Himalia 1, so developers using the SDK get this protection without changing their integration. The specialist models extend coverage to situations where a single general model falls short, such as coding agents that read and write code and security agents that act on their own. Together, the family helps teams catch malicious instructions hidden in user input, tool outputs, and agent context before they cause damage. Built for developers and security teams shipping agent-based applications who need protection that fits into existing frameworks and SDKs, including Next.js, Express, FastAPI, and Node.js. Use it to safeguard chatbots, coding assistants, and autonomous agents against prompt injection and related abuse.