Anthropic model sent false murder tip to Philadelphia police

Anthropic says its Claude Haiku 4.5 model generated a false homicide tip during random webpage testing and submitted it to Philadelphia police on July 18. Police flagged it as spam and never forwarded it to investigators, stressing human vetting is required.

• Anthropic: Claude Gave False Murder Tip to Police2026-10-12
• Anthropic Cuts Live Internet Access for Internal AI Tests After Claude Exploits Injection Flaws2026-10-10
• Philadelphia police say Anthropic model filed false murder tip2026-10-09
• Anthropic Releases Claude Haiku 5.52026-10-07
• OpenAI, Anthropic tell UK lawmakers to regulate AI2026-10-13