Philadelphia:An artificial intelligence model developed by Anthropic mistakenly submitted a false tip about an unsolved homicide to the Philadelphia police. The company informed authorities of the incident nearly three months later, as reported on Friday.
According to Anadolu Agency, the erroneous submission took place on July 18 during an automated test involving randomly selected websites, as reported by The Wall Street Journal. The AI model submitted the tip through PhillyUnsolvedMurders.com, posing as someone with information about an unsolved case.
The Philadelphia police stated that the submission was immediately categorized as spam and was never forwarded for investigation. They expressed concern over the delay, highlighting it as unacceptable that it took the company two months to detect and report the incident. Anthropic discovered the error on September 28 and notified the police on October 7.
In response, Anthropic terminated the automated testing process responsible for the submission and implemented an additional validation mechanism for future tests. The incident underscores the importance of preventing technology systems from submitting false information to law enforcement, particularly in cases involving real victims and investigations.
The police department emphasized the importance of the platform, which lists unsolved homicides and accepts public tips, alongside a $20,000 reward for information leading to an arrest. Authorities continue to encourage the public to submit legitimate information through the platform. Anthropic did not respond to requests for comment on the incident.