Reporting an AI product that behaves badly or generates harmful output has been a fragmented, often futile process. Now, a group of AI researchers has launched FLARE-AI (Flaw Reporting for AI), a crowdsourced website designed to provide a centralized system for reporting and tracking AI harms. According to WIRED, the open-source platform allows anyone to report issues such as a chatbot generating malware, leaking personal information, or triggering delusional thinking in users.
How FLARE-AI Works
The platform, described as similar to Downdetector for AI, enables users to submit reports that can be verified by others. The code behind FLARE-AI is open source, allowing external validation and routing of reports to model makers and organizations like MITRE, a nonprofit that tracks problems with technical systems. The system was developed in collaboration with 49 AI experts from 32 different organizations, according to a paper outlining the initiative.
"Right now, there is no centralized, accountable way to report flaws in AI systems," said Avijit Ghosh, an artificial intelligence policy researcher at HuggingFace who co-led development of FLARE-AI with computer scientists Elaine Zhu and Shayne Longpre. Ghosh noted that problems with AI systems span psychological harm, discrimination or bias, and misinformation. Different companies maintain different standards around such issues, meaning some problems go unrecognized. "In the absence of a coordinated disclosure system, there are no external mechanisms to enforce transparency," he added.
The Need for Standardized Reporting
Jessica Ji, a researcher at the think tank Center for Security and Emerging Technology, endorsed the initiative. "I think it's a really good initiative," she said, noting that existing reporting mechanisms are fragmented and that AI models are black boxes. "I'm in support of anything that makes AI more transparent."
Recent incidents highlight the ease with which AI can go wrong. LayerX disclosed a method to dupe AI-infused web browsers, including OpenAI's Atlas and Perplexity's Comet, into bypassing their guardrails—for example, convincing the AI that it was playing a game could lead to it attempting to hack a website. (LayerX says the affected companies have since fixed the issue.) In April, security researcher Johann Rehberger discovered a way to trick Claude into divulging personal data using images generated by ChatGPT. Last year, OpenAI updated its models after discovering they were overly sycophantic, sometimes appearing to encourage delusional thinking.
| AI Harm Category | Example from Source |
|---|---|
| Malware generation | Chatbot generates malicious code |
| Psychological harm | Triggering delusional thinking in users |
| Privacy violation | Leaking personal information |
| Guardrail bypass | AI-infused browsers tricked into hacking |
| Misinformation | Overly sycophantic responses encouraging false beliefs |
Challenges and Government Support
Rumman Chowdhury, CEO and founder of Humane Intelligence PBC, noted that FLARE-AI could be a useful tool for AI developers but warned of serious challenges. One is managing a flood of reported issues, many of which may not be serious. Another is ensuring reporting schemes are backed by credible and authoritative organizations. Last month, a congressional bill introduced by Representatives Deborah Ross could provide such backing, proposing a central US government role in tracking AI misbehavior. Members of the FLARE-AI group consulted on that bill, according to WIRED.
Implications for Enterprise AI
The initiative arrives as organizations increasingly deploy AI in critical business functions—from supply chain optimization to trade finance. For CTOs and technology leaders, a standardized flaw-reporting mechanism offers a way to stay informed about vulnerabilities in AI tools they may rely on. The researchers argue that their system could prove crucial as AI is adopted more widely and as agentic systems gain greater power. For enterprises, the ability to quickly report and verify AI issues could reduce operational risk and accelerate vendor accountability.
The FLARE-AI platform is now live, and its open-source nature invites broad participation. As Ghosh put it, without external enforcement, transparency will remain elusive. This initiative, backed by researchers and potentially the US government, aims to fill that gap.
Sources:
