Hi Product Hunt, I built La AutopsIA from daily work with AI systems and from a simple need: having one place where real AI failures can be documented, classified and compared over time. La AutopsIA is a public observatory of documented AI incidents. It tracks hallucinations, bias, security alerts, vulnerabilities and operational failures, and organizes them with categories, severity, affected countries, benchmarks, an AI model comparator, public exports and monthly datasets. A key part of the project is the monthly snapshot system. Every month, La AutopsIA publishes a fixed version of its datasets, so the state of the observatory can be cited, compared and preserved over time. The June 2026 snapshot is already published across Zenodo, Hugging Face and Kaggle, with a permanent DOI on Zenodo. I use AI every day. This project is not written from outside the field. It comes from using these systems, building with them, and documenting where they fail in real cases. I am launching it here to get honest feedback on the observatory, especially: • whether the product is clear • whether the filters and categories are useful • whether the AI comparator is easy to understand • whether the monthly dataset snapshots are valuable • what export or dataset formats would make it more useful Thanks for taking a look.

𝐋𝐚 𝐀𝐮𝐭𝐨𝐩𝐬𝐈𝐀

Real AI failures, incidents and security alerts

Real AI failures, incidents and security alerts