Arcjet has added AI prompt injection protection to its runtime policy engine, using a purpose-trained LLM to detect and block malicious prompts before they reach embedded AI models. Built on WebAssembly, the engine integrates with JavaScript, Python, and frameworks like LangChain and Vercel AI SDK, adding under 100ms of overhead. Initially covering prompt-extraction and shell-injection attacks, the capability aims to prevent attackers from overriding application behavior, exposing hidden prompts, or extracting data at runtime rather than after the fact via log analysis.
Sort: