Anthropic’s artificial intelligence gave a false homicide tip to Philly police, triggering a meeting with the company
The incident happened on July 18 but was not discovered by the Silicon Valley company until September, authorities say.
An artificial intelligence model operated by Anthropic submitted false information about an unsolved Philadelphia homicide to the police department’s public tip forum this summer, law enforcement officials said Friday.
The unprecedented incident occurred in July and raised concern among department leaders, who met with Anthropic representatives to discuss the interaction on Thursday, police said.
Sgt. Eric Gripp, a police spokesperson, said an Anthropic AI model contacted the department’s tip forum, PhillyUnsolvedMurders.com, around 11:30 p.m. on July 18.
The model gave information purporting to come from someone with information about an unsolved homicide, he said.
It was not immediately clear which homicide the AI told the department it had information about.
The email was flagged as spam and remained in that folder until Wednesday, when Anthropic alerted police to the incident, authorities said. No city or police data was accessed as a result of the interaction, Gripp said.
Anthropic told police the false tip was generated as the company tested interactions between its AI model and “randomly selected websites,” according to Gripp.
A spokesperson for Anthropic did not immediately return a request for comment.
The company discovered the error on Sept. 28. As a result, Gripp said, Anthropic terminated testing for that model and added safeguards for further trials.
The news arrives ahead of the release of a report from Anthropic on Friday that is expected to detail the Philadelphia police incident as well as other instances of unintended behavior from its AI models.
Earlier this summer, the company made international headlines when it announced that its Claude AI models had broken free from isolated testing environments and, in several instances, hacked into the internal infrastructure of outside organizations.
Anthropic notified those companies, whose names it did not disclose, about the breach on July 27.
A little more than a week earlier, authorities said, Philadelphia police received the artificially generated homicide tip.
This is a developing story and will be updated.
