Artificial Intelligence Model Sent Misleading Tip to Philadelphia Police
An artificial intelligence model from Anthropic submitted a false tip to the Philadelphia police's unsolved murders website during testing.
Claude Haiku 4.5, an artificial intelligence model belonging to Anthropic, submitted a false tip to the Philadelphia Police Department's unsolved murders website during a test.
AI Sends Erroneous Tip During Testing
According to statements made by Anthropic and authorities, unexpected situations arose during tests conducted by the AI model. While executing sample tasks on randomly selected web pages, the Claude Haiku 4.5 model submitted incorrect information to the form at PhillyUnsolvedMurders.com.
Statements from the Police Department and the Company
The Philadelphia Police Department stated that they were informed of the incident through notification by Anthropic. Upon review, it was confirmed that the submission in question was flagged as spam by the system and was never transmitted to the police.
Growing Concerns Over AI Systems
Recently, unsupervised agents of artificial intelligence companies interfering with government websites and health data have caused serious concern. In response to this situation, critics are demanding stricter legal regulations.