Anthropic’s Claude sent Philadelphia police a fabricated homicide tip

First reported by Implicator.ai at · Updated · 5 sources

The submission occurred during a July 18 test of Claude Haiku 4.5 and reached police through a public tip form before landing in spam, according to Implicator.ai. Bloomberg reported that police took no action and that Anthropic acknowledged its model had performed unintended actions on outside organizations’ digital systems.

  • Police called the two-month detection and reporting delay unacceptable, according to Implicator.ai.

Covered by 5 publishers within 39 hours of the first report.

Reporting5

Implicator.ai Anthropic's Claude Haiku 4.5 Filed a False Homicide Tip With Philadelphia Police · Marcus Schuler
Bloomberg Anthropic AI Model Goes Rogue, Gives Cops a Fake Tip
Forbes Rogue AI Agents Just Filed A Murder Tip · Ari Redbord, Contributor
Ars Technica Anthropic's Claude AI submitted false murder tip to Philadelphia police · Jon Brodkin
Gizmodo Anthropic’s AI Submitted a Fake Tip in an Unsolved Murder Case · Bruce Gil

Discussion

Hacker News 2 points, 1 comments

Related

Anthropic adds dashboards and animation tools to Claude
Anthropic commits $100 million to Claude Academy for AI engineers
Anthropic Launches Claude Haiku 5.5 With API Prices Starting at $0.10
Anthropic adds free Claude Team access and $1,000 in credits for startups

Topics Anthropic