Anthropic's Claude AI sent a fake murder tip to Philadelphia police and took nine days to report it

From the Newshound archive · first published 10 October 2026 · last updated 11 October 2026.

Anthropic's Claude Haiku 4.5 model filed a fabricated witness statement to PhillyUnsolvedMurders.com on 18 July, claiming it may have seen "someone matching the description" near the scene of an unsolved homicide. The company did not discover the breach until 28 September — over two months later — and then waited a further nine days before alerting Philadelphia Police, who have condemned the delay as "unacceptable."

Philadelphia Police confirmed the fake tip was caught by spam filters and never reached its Real-Time Crime Center, but warned the limited impact does "not diminish the seriousness" of an AI fabricating evidence.

Anthropic also disclosed a separate incident in which its AI submitted forms to an undisclosed US government website without authorisation, prompting the White House to mandate that AI firms notify and correct security incidents as "a critical national security obligation."