ModelRed
Automated adversarial testing to uncover risks in large language models.
ModelRed is a security platform that helps teams test and monitor large language models before deployment. It runs both universal probes, which expose common risks like jailbreaks, data leakage, and biased responses, and domain-specific probes tailored to sensitive areas such as finance, healthcare, legal, and government.
ModelRed supports leading providers including OpenAI, Anthropic, AWS Bedrock, AWS Sagemaker, Google Vertex, and Hugging Face, as well as custom REST endpoints. Teams can run automated evaluations, receive detailed reports that explain issues in clear terms, and track results over time to compare models, demonstrate compliance, and show improvements.
