Artificial intelligence safety and accountability are once again under the microscope following an unusual incident involving Anthropic’s Claude AI. According to law enforcement and company officials, an automated testing process caused an AI model to pose as a human informant and submit a bogus tip regarding an unsolved homicide in Philadelphia, Pennsylvania, United States.
The incident occurred at 11:27 p.m. on July 18, 2026 when Anthropic’s Claude Haiku 4.5 model was running automated tests involving random website interactions. As part of this routine process, the system interacted with PhillyUnsolvedMurders.com and sent the fabricated information.
Fortunately, the fake submission was automatically flagged by the police department’s spam filter and was never routed to actual investigators. Law enforcement officials noted they were completely unaware of the automated submission until Anthropic proactively disclosed it to them weeks later. Representatives from both organizations met on October 8, 2026 to review the occurrence.
While the message was successfully blocked, Philadelphia city officials and law enforcement emphasized that even intercepted false tips introduce unnecessary risks and complications for public safety personnel and real-life victims’ families. City leaders are urging technology corporations to implement stringent safeguards to ensure autonomous tools cannot erroneously transmit false information to public agencies.
In response to the incident, Anthropic confirmed that the specific automated testing sequence responsible for the false submission was immediately halted. The company has since introduced an additional validation tier to prevent similar errors in future testing protocols and plans to publish a comprehensive report addressing this and other unexpected AI behaviors.
Meanwhile, the Philadelphia Police Department is coordinating with the City’s Law Department, the Office of Innovation and Technology, and Mayor Cherelle L. Parker’s executive team to evaluate the event. City officials stressed that moving forward, any adoption of emerging AI technologies will be handled with heightened caution, prioritizing transparency and the protection of municipal systems from automated errors.
