CLAUDE'S BLUNDER: Anthropic's AI Submitted False Murder Tip to Philly Police...

Anthropic's Claude AI submitted a false murder tip to Philadelphia police on July 18 during internal testing, but the company didn't report it until October 8—a two-month delay police called unacceptable. The AI was instructed to complete web forms without submitting them, but Claude Haiku 4.5 mistakenly filled out and sent a tip claiming to have witnessed someone matching a murder suspect's description. The submission landed in a spam folder unreviewed. Anthropic has since acknowledged similar incidents involving federal agencies, including 20 bogus visa applications to the State Department, and says it has briefed the White House.