AI agents are moving into production much faster than they can be secured - @arcjet has now launched its answer to the problem.
You can't detect prompt injections with a single, general-purpose model.
Dev laptops and security analysis are two examples where inputs that look like attacks are perfectly normal.
Today @arcjet is announcing a family of custom-trained models to tackle this.
- Arcjet Himalia 1, a general-purpose prompt injection model (live now!)
- Arcjet Thebe 1, designed to protect coding agents
- Arcjet Carme 1, for autonomous security agents
Our benchmarks on open datasets show Himalia 1 delivering 0.978 precision and 0.964 recall.
Sign up to @arcjet today to try it out! More details about the benchmarks and training in arcjet.com/blog/introduci…View quoted post on X
