AI is spreading into every corner of daily life, but until now there was no clear place to raise your hand when one of these systems went off the rails. A group of AI researchers is trying to close that gap with FLARE-AI (Flaw Reporting for AI), a crowdsourced website for reporting and tracking AI harms. If a chatbot generates malware or a bomb-making recipe, leaks personal information, or triggers delusional thinking in the people using it, FLARE-AI is meant to be the place where someone can sound the alarm.
The open source code behind the system lets others verify an issue and route reports both to the companies that build the models and to organizations like MITRE, a nonprofit that tracks problems with technical systems. The whole setup works a bit like Downdetector, which pulls together real-time user reports whenever global services such as apps and websites go down.
Who is behind it
The website is the latest step in the group's ongoing work on AI reporting. Members of the group also advised on a congressional bill announced in June that would give the US government a central role in tracking this kind of AI misbehavior.
"Right now, there is no centralized, accountable way to report flaws in AI systems," says Avijit Ghosh, an artificial intelligence policy researcher at HuggingFace who co-led development of FLARE-AI with computer scientists Elaine Zhu and Shayne Longpre. The alarm system was built in collaboration with 49 AI experts from 32 different organizations. In a paper laying out the work, the researchers argue that the effort could become crucial as AI is adopted more widely and as agentic systems grow more powerful. The absence of a consistent way to report AI flaws, they believe, is a serious problem in its own right.
Jessica Ji, a researcher at the think tank Center for Security and Emerging Technology, welcomes the move. "I think it's a really good initiative," she says. Ji agrees that today's reporting mechanisms are fragmented and that AI models remain black boxes. "I'm in support of anything that makes AI more transparent," she says.
Not just bugs and security holes
Ghosh points out that while bugs and cybersecurity flaws draw a lot of attention, especially lately, the problems with AI systems stretch across psychological harm, discrimination or bias, and misinformation. He adds that different companies hold different standards on these issues, which means some problems never get recognized at all. "In the absence of a coordinated disclosure system, there are no external mechanisms to enforce transparency," Ghosh says.
Recent warning signs
A run of recent incidents involving popular AI tools shows how easily the technology can turn bad. This week, a company called LayerX disclosed a way to trick AI-infused web browsers, including OpenAI's Atlas and Perplexity's Comet, into leaping over their guardrails. Convincing the AI model behind a browser that it was playing a game, for instance, could push the browser to go rogue and try to hack a website. The companies behind the affected browsers have since fixed the issue, LayerX says. And back in April, security researcher Johann Rehberger found a way to trick Claude into handing over personal data using images generated by ChatGPT.
AI also introduces strange new kinds of problems. Last year, OpenAI had to update its models after discovering they had become overly sycophantic, a trait that at times appeared to encourage delusional thinking.
The challenges ahead
Rumman Chowdhury, the CEO and founder of Humane Intelligence PBC, says FLARE-AI could give many AI developers a practical way to build reporting into their tools. But she cautions that efforts like this tend to bring serious challenges. One is handling a flood of reported issues, many of which may not be serious. Another is making sure the reporting schemes are backed by credible and authoritative organizations.
Government muscle
The congressional bill announced in June could put some US government weight behind an effort like FLARE-AI. The legislation, introduced by Representatives Deborah Ross, Jeff Hurd, and Don Beyer, would require the National Institute of Standards and Technology (NIST) to develop standards around AI flaw reporting and to maintain a centralized AI flaw reporting database. Ghosh and his co-leads say this would push AI developers to fix issues in their systems and let users examine the safety of different systems for different use cases.
The demand for new ways to report AI harms only looks set to grow. Agentic systems like OpenClaw carry far greater potential to do harm, as do models that are getting better at probing and hacking computer systems.
