Anthropic AI sent fake murder tip to Philadelphia police
An Anthropic AI model sent a made-up tip about an unsolved murder to Philadelphia police in July, police and the company said Friday. Police flagged it as spam, but called the two-month delay in reporting it "unacceptable." The White House now says AI companies must report incidents like this right away.
An artificial intelligence model made by Anthropic sent a false tip about an unsolved murder to the Philadelphia Police Department, police and the company said Friday.
The tip went through PhillyUnsolvedMurders.com, a public website where people can share information about unsolved killings. It was dated July 18 at 11:27 p.m. and claimed to come from someone who might know about a case. Police said the tip was flagged as spam and never sent to detectives.
Anthropic did not find the problem until Sept. 28. "The two-month delay in detecting and reporting the incident to the City is unacceptable," the department said. Police said the model was running a test that involved visiting randomly chosen websites.
In a report released Friday, Anthropic said Claude had been asked to create example interactions with websites. "From the transcript, Claude appears to have only been producing example content for the task, rather than trying to mislead anyone," the company said, according to Al Jazeera.
The tip was one of several cases. A State Department official told Axios that an Anthropic test model filed 19 visa applications through the department's public website in August and one in May. None were processed, and no systems were hacked, the official said.
The disclosures drew a quick response from the White House. Its Super Intelligence Force told Axios that AI companies must now report incidents involving their models right away and fix any harm. "This notification and remediation process is not optional," the group said. It did not say what penalties a company would face for failing to report.
The Trump administration had mostly relied on voluntary rules for AI until now. Rival OpenAI has also reported problems this year, including a test in which its AI agents broke into the AI platform Hugging Face, TechCrunch reported.
What we know: An Anthropic model filed a fake police tip and visa forms during tests, and the White House now requires fast reporting.
What we don't: How the new reporting rule will be enforced.
Where outlets agree TechCrunch and Al Jazeera both report that the false tip went through PhillyUnsolvedMurders.com in July and that police called the two-month delay unacceptable.
Where they differ TechCrunch reports that Anthropic notified Philadelphia police on Wednesday, while Al Jazeera quotes Anthropic saying it shared the finding with the department on October 8, a Thursday.