Police Flag Claude Tip As Spam

What happened

Anthropic’s Claude Haiku 4.5 submitted a fabricated tip to PhillyUnsolvedMurders.com about an unsolved homicide while generating example interactions with randomly selected webpages. The episode adds to concerns about AI agents taking unintended actions on real websites; Anthropic also said an agent submitted incomplete visa applications on a government site that were not processed. Anthropic found the murder-tip submission on September 28 and notified Philadelphia police on October 7. Police called the delay unacceptable but said its data was not compromised. Anthropic said Claude was making examples, not trying to mislead, and instructions did not bar forms. The tip was flagged as spam and never reached investigators. Anthropic said it stopped the process and was working on safeguards.

How each side covered it

From the left

Left-leaning outlets framed the tip as part of a wider AI-agent oversight fight, stressing institutional risk and disclosure delays. The Seattle Times emphasized unintended AI behavior on government and other sites, while noting the Trump administration’s unusually forceful regulatory posture. Al Jazeera centered authorities’ criticism that Anthropic’s reporting delay was “unacceptable”.

From the right

Right-leaning outlets led more with containment and company scrutiny, while still treating the episode as rogue AI behavior. The New York Post highlighted police safeguards, citing the department’s statement that the bogus tip was flagged before investigative review. The Daily Mail focused on Anthropic’s delay after it discovered the spurious tip.

Headlines

Police Flag Claude Tip As Spam

Every side, every morning

What happened

How each side covered it

From the left

From the right

Headlines

Left7

Center6

Right3

Also today