A group of AI researchers has launched a crowdsourced website called FLARE-AI to let users report and track harmful AI behaviors. The platform allows users to flag issues like AI models generating malware, bomb-making recipes, or leaking personal information. The open-source code enables others to verify reports and route them to model developers and organizations like MITRE. This system is similar to Downdetector, which compiles real-time user reports for service outages.
FLARE-AI is part of ongoing work by the researchers, who also consulted on a congressional bill announced in June that would have the US government track AI misbehavior. Avijit Ghosh, an AI policy researcher at HuggingFace, says there is currently no centralized way to report AI flaws. The website was developed in collaboration with 49 AI experts from 32 organizations. The researchers argue that their initiative is crucial as AI adoption grows and agentic systems gain more power.
The lack of a consistent way to report AI flaws is a significant problem, according to the researchers. Jessica Ji, a researcher at the Center for Security and Emerging Technology, supports the initiative, noting that existing reporting mechanisms are fragmented and AI models are black boxes. Ghosh adds that AI issues span psychological harm, discrimination, and misinformation, with varying standards across companies. The absence of a coordinated disclosure system means no external mechanisms to enforce transparency.
Source: wired
