We tested 9 AI models with 37K+ security tests

Show HN: We tested 9 AI models with 37K+ security tests

We tested 9 AI models with 37K+ security tests is a company in show hn launch.

Profile

Hi everyone,We built ModelRed to test AI models and apps for security issues. Ran 4,182 attack probes against 9 leading models to see what would break.Leaderboard: https://modelred.ai (no signup, just check it out)Claude scored 9.5/10 but still failed on medical/financial prompts. Mistral Large scored 3.3/10. The gap between best and worst is huge.We test for prompt injections, data leaks, jailbreaks, risky tool calls, domain specific hacks, basically everything that goes wrong when your LLM has access to real data and APIs. The platform runs these tests continuously and blocks CI/CD deployments when scores drop.Works with any provider (OpenAI, Anthropic, AWS, Huggingface endpoints, OpenRout

Within the show hn launch space, the company focuses on its core offering rather than a broad product line.

Directory data places it at the Show HN launch stage.

Website: www.modelred.ai/leaderboard