You Can Now Sound the Alarm on AI Behaving Badly

FLARE-AI, or Flaw Reporting for AI, is a new crowdsourced website created by AI researchers to address the lack of a centralized system for reporting issues with artificial intelligence systems. Currently, different companies have varying standards for handling AI flaws like bias, misinformation, or security vulnerabilities, making it difficult to consistently disclose serious problems. This new platform allows users to report a wide range of AI misbehaviors, including instances where chatbots generate malware, leak personal information, or induce delusional thinking. The system is designed to facilitate community verification of reported issues through open-source code and then routes these verified reports to AI model makers and organizations that track technical problems, such as MITRE. The project aims to improve transparency and accountability in the rapidly expanding AI industry by providing a consistent framework for documenting and addressing AI flaws. This initiative aligns with a proposed US congressional bill that would mandate a centralized database for AI flaw reporting, highlighting the growing recognition of the need for such a system.