Skip to content
Bramble

Technology

Philadelphia police say an Anthropic AI model submitted a false homicide tip

Philadelphia police said an Anthropic AI model submitted a false tip about an unsolved murder through a police tip website. Anthropic notified the department on Oct. 7 and says an automated testing process was responsible.

4 outlets · 2L · 2C · 0R First reported Account updated
Image: CBS News
Image: CNA
Image: TechCrunch
Image: The Verge

1 / 4

The story, neutrally told

Mixed · 2Philadelphia police said a "false homicide tip" about an unsolved murder, submitted through the PhillyUnsolvedMurders.com website, was generated by an Anthropic artificial intelligence model. Left · 1The website is an online platform the department launched to gather tips on unsolved homicide cases. Centre · 1TechCrunch, citing 6abc Action News, reported that the tip was submitted on July 18, that Anthropic did not discover it until September 28, and that police had not seen it because it was marked as spam.

Mixed · 2Anthropic notified the department on Oct. 7 and met with it the next day, according to TechCrunch. Centre · 1According to the police, Anthropic said an automated testing process was responsible for the submission, that the process was stopped once the incident was uncovered, and that an additional validation mechanism has been introduced for future testing. Left · 1Anthropic told police it plans to publish a report on Friday describing what happened and "other instances of unintended model behavior"; the department said it released the information ahead of that "in the interests of full government transparency and accountability."

Centre · 1The department told 6abc that Anthropic "must strengthen its safeguards to prevent similar incidents from impacting city systems without the city’s knowledge" and called the two-month delay in detecting and reporting the incident "unacceptable." Centre · 3Anthropic did not immediately respond to requests for comment from CBS News, CNA or TechCrunch.

Every sentence links to the reporting it rests on. The pill in front of each says where its sources sit: Left, Centre or Right when one side supplies at least half of them, Mixed when they are evenly split. The number is how many outlets it cites.

Left2 outlets

Framing
Short report built on the police statement, leading with Anthropic's notification and planned report.
Emphasis
Police transparency and Anthropic's forthcoming report.
Leaves out or plays down
Does not mention the July 18 date, the spam filtering, or Anthropic's testing-process explanation.
Charged language
“false homicide tip”
For example
“Philadelphia Police are providing this information to the public ahead of that publication in the interests of full government transparency and accountability,"” — CBS News

Centre2 outlets

Framing
CNA gives a brief account with Anthropic's explanation; TechCrunch adds the timeline and uses the incident to warn about autonomous AI agents.
Emphasis
The testing-process explanation (CNA); the two-month delay and risks of unsupervised AI (TechCrunch).
Leaves out or plays down
CNA omits the dates and the police criticism. TechCrunch omits Anthropic's testing-process explanation and the planned report.
Charged language
“unacceptable”“unchecked access”
For example
“this incident highlights the danger of giving AI the ability to carry out tasks without any human supervision.” — TechCrunch
“said that an automated testing process was responsible for the submission, according to the police.” — CNA

Right0 outlets

No right outlet in our sources has covered this story yet.