Anthropic delayed two weeks before correcting AI's false murder tip to Philadelphia police
Anthropic took two weeks to notify Philadelphia police that its AI system had submitted a hallucinated lead regarding an unsolved murder case.
Key points
- Anthropic's AI submitted a false tip about an unsolved murder to Philadelphia police.
- The company took two weeks to notify authorities that the lead was hallucinated.
- The incident highlights risks of AI systems generating fabricated information for law enforcement.
Anthropic waited two weeks to inform Philadelphia police that its artificial intelligence system had submitted a fabricated lead concerning an unsolved murder. According to a report by TechRadar, the delay between the initial false submission and the correction raises concerns about the reliability of automated intelligence feeds used by law enforcement agencies.
The AI system generated a hallucinated tip, meaning it created information that did not exist in reality. This false lead was sent directly to the Philadelphia police department. Anthropic eventually notified the police that the information was incorrect, but the two-week gap suggests a lack of immediate verification protocols for high-stakes outputs.
Context and Background
TechRadar reported that Anthropic took two weeks to notify Philadelphia cops that its AI had sent them a hallucinated lead. The incident involves an unsolved murder case, where the AI provided a tip that appeared credible but was entirely fabricated. This type of error is known as hallucination, where large language models generate plausible-sounding but factually incorrect statements.
The delay in correction is significant for law enforcement operations. Police departments rely on timely and accurate information to solve crimes. A false lead can divert resources and attention away from genuine investigative avenues. The two-week period during which the police may have acted on or considered the false tip represents a substantial window of potential operational inefficiency.
Who Is Affected
Philadelphia police are the primary entities affected by this incident, as they received the false information directly. Law enforcement agencies that integrate AI tools into their investigative workflows face similar risks. The incident underscores the vulnerability of relying on automated systems for critical intelligence without human verification.
Security professionals monitoring AI deployment in public safety sectors are also impacted. This event serves as a case study for the potential consequences of AI errors in sensitive environments. It highlights the need for robust oversight mechanisms when AI systems interact with law enforcement databases or communication channels.
What should you do?
Organizations using AI for critical decision-making should implement strict human-in-the-loop verification processes. Law enforcement agencies should treat AI-generated leads as unverified until corroborated by traditional investigative methods. Continuous monitoring of AI outputs for hallucinations is essential to prevent resource misallocation.
Background: Large language models
Large language models are a specific type of neural network trained on text. Generative AI is the broader umbrella term for any system that creates new content, including images and code. Not all generative AI uses large language models, but most modern text generators do.
What to do and how to stay safe: Anthropic
- Implement mandatory human verification for all AI-generated intelligence before acting on it.
- Establish clear protocols for correcting false information submitted by AI systems immediately.
- Monitor AI outputs for signs of hallucination, especially in high-stakes scenarios like criminal investigations.
- Review integration practices between AI tools and law enforcement communication channels to ensure accuracy.
Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
How long did Anthropic take to correct the false murder tip?
Anthropic took two weeks to notify Philadelphia police that the AI-submitted lead was hallucinated.
What type of error did the AI commit?
The AI committed a hallucination, submitting a fabricated lead about an unsolved murder to the police.
Which police department received the false information?
The Philadelphia police department received the false tip regarding the unsolved murder case.
