In Philadelphia, the Anthropic AI Model sent a False Murder Message

Philadelphia police received a false message generated by the Anthropic artificial intelligence model through a website for collecting information about unsolved murders. This was reported on October 9 by CBS News TV channel with reference to the police.

"Philadelphia police reported that there was a "false murder report" that came through the website. PhillyUnsolvedMurders.com , was generated by the Anthropic artificial intelligence model," the article says.

According to Anthropic, the message was created by the Claude Haiku 4.5 model as part of automated testing. The AI generated a text on behalf of a person who allegedly could have information about the murder under investigation, but left the fields with the name and contact information blank.

The company explained that Claude was tasked with generating and executing examples of actions on random web pages. At the same time, the model did not receive instructions to access the police website or send malicious data, but the instructions did not exclude the possibility of filling out and sending forms.

After the incident was discovered, Anthropic stopped automated testing, during which the message was sent. The police stressed that there were no signs of unauthorized access to its systems or leakage of departmental data.

Agent under cover: the largest player in the IT market has invested 7.5 billion in countering AI threats

Why did key Russian technology companies simultaneously begin to create protection against neural network solutions?

On September 26, The New York Times reported that OpenAI artificial intelligence agents had attempted unauthorized access to the Internet portals of three US government departments. Software agents tried to hack the official website of the US Department of Education to collect data, extracted official information from the database of the Census Bureau using other people's accounts found on the Network.