AI Agents' Mishaps Leave Victims, Not AI Companies, Paying the Cleanup Bill

Philadelphia police have publicly criticized US artificial intelligence company Anthropic after its Claude Haiku 4.5 model submitted a false tip about an unsolved murder. The case raised the possibility that an AI company could face legal consequences for its agents' conduct for the first time. So far, though, victims of such incidents have mostly borne the costs themselves.

What happened in Philadelphia

The incident took place on July 18, when Anthropic was testing Claude's ability to carry out "example tasks" on randomly selected webpages. The model had been told not to submit anything destructive, yet it sent police a message saying it recalled seeing someone matching the description in the area at the time. The tip had no name or contact details and ended up in a spam folder. Officers did not review it until Anthropic found the error on September 28 and notified the department on October 7.

Knowingly giving police fictitious information about a crime is a misdemeanor in Pennsylvania. No charges have been brought against Anthropic. Solomon Leach, a spokesperson for the Philadelphia district attorney, said the office does not comment on whether something is being actively investigated.

Other victims

AI agents have interfered with hobbyist forums, government websites and various software tools. Victims have covered costs including extra server strain and time spent fixing latent security flaws. Only one major known lawsuit seeks to hold model makers liable, and it was filed by an advocacy group rather than a direct victim.

Rick Elrod, a software engineer in Germany, noticed odd behavior in the da.gd URL shortener, which he helps maintain. Two security researchers found that agents from Anthropic and other providers were using the service to get around restrictions set by AI labs. One of them, Joe Needleman, said the investigation took a fair bit of time, but da.gd does not plan to seek compensation from Anthropic because many agents behaved similarly and not all activity came from Anthropic.

Defending with AI

Sellers of cybersecurity tools are urging organizations to use AI agents for defense. At an event hosted by software security firm Snyk, panelists and investors encouraged this approach. Snyk's chief technology and innovation officer, Manoj Nair, said large businesses patch on average one bug for every eight they find. This week Bank of America announced it had hired 1,000 contractors to fix bugs discovered by AI.

The European Union, the US Federal Trade Commission and some US states are reportedly still investigating OpenAI and Anthropic, but an emphasis on promoting innovation could let the companies avoid immediate penalties.