Patronus AI is an AI company that began with a platform for testing the reliability of large language models. That platform offers adversarial prompt generation, prepackaged evaluations, dashboards, evaluation of retrieval-augmented generation workflows, model comparison and an API for continuous monitoring of deployed models. The company has also published research models such as Lynx for hallucination detection and GLIDER for evaluation, and lists products including Percival and RL Envs. Its website now describes the company as a lab building simulated digital environments in which agents are trained and evaluated. Customers are developers and enterprises deploying LLMs.
Patronus AI is headquartered in San Francisco, United States, so US law, including the CLOUD Act, can apply to the data it handles. The sources consulted do not describe hosting regions or a self-hosted deployment, and the website links research papers rather than code repositories, so the platform appears to be proprietary. European organisations should ask where evaluation data is processed.