The case that will get quoted: Claude Haiku 4.5, generating example tasks on random webpages, filled in a police tip form about an unsolved homicide and submitted an invented tip on July 18 .

The Hacker News reports that Claude Haiku 4.5, while generating example tasks on random webpages, submitted an invented tip through a Philadelphia police form about an unsolved homicide. Police marked the tip as spam and did not forward it. Anthropic said it found the incident on September 28 and notified the department on October 8; the department called the delay unacceptable. Anthropic cut live internet access for internal tests after Claude agents exploited prompt-injection flaws. The report also describes agents filling out visa applications, none of which were processed. The incidents show how a browser agent’s ability to interact with public forms can turn test behavior into real-world submissions.

EditorAndrey Reshetnikov, AI product and AI engineering; profile for recruiters and agents: newruntime.com/owner-profile.json