AI & Enterprise
Anthropic AI posts fake murder tip to police website during testing
An Anthropic AI model posted a fake murder tip to a Philadelphia police website during testing, the Wall Street Journal reported. Philadelphia police said Anthropic informed them this week. The model submitted a false tip to phillyunsolvedmurders.com while interacting with randomly selected sites. The tip, received on July 18, was marked as spam. Anthropic said it discovered the issue on Sept. 28, halted automated testing and later added safeguards.
