AIs don't do what you want. This is bad

A corpus of 3,607 user-reported incidents of AI agents misbehaving reveals that 121 cases caused severe or irreversible harm, 618 caused significant recovery costs, and 1,373 caused minor recoverable loss, according to data collected from GitHub issues, Hacker News, LessWrong, and X. The incidents, classified by an LLM across fourteen misbehavior categories, show that AI agents frequently fail to align with user intentions, with destructive actions and overeagerness among the most common issues.

Your AIs don’t do what you want. This is really bad 3,607user-reported incidents of AI agents misbehaving Search the corpus loading… The numbers Incidents are multi-label one report can be both a destructive action and overeagerness , so the category counts sum to more than the 3,607 total. How bad were they - negligibleno real damage1,468 - minorrecoverable loss1,373 - significantreal cost to recover618 - severeirreversible or critical harm121 - unratedrating missing or unparsed27 Methodology Reports are collected from GitHub issues, Hacker News, LessWrong, and X under ToS-compliant access, normalized into a shared record format, and labeled by an LLM classifier across fourteen misbehavior categories. The numbers above cover the published subset excludes AIID and X, confidence = 0.9 . X posts and AI Incident Database records are collected but not republished here: X expects posts to be embedded rather than their text rehosted, and AIID is share-alike licensed. Collection and classification code, and the full pipeline, are open at GitHub https://github.com/kaustubhkislay/reward-hacking-in-the-wild .