An Anthropic AI model sent a false tip about an unsolved homicide to the Philadelphia Police Department (PPD), and the company didn’t find out for more than two months. The PPD reported that the tip came through PhillyUnsolvedMurders.com on July 18th, according to a report from local station 6abc, as noted by The Verge.
What We Know So Far
The PPD released a statement on Friday saying the tip was submitted by an Anthropic AI model through the department’s online crime tip system. A tipline allows the public to report information about crimes, providing detectives with leads from real witnesses. Unfortunately, this time, the information was false.
According to Engadget, the false tip didn’t seem to divert police resources. TechCrunch reports that Anthropic only discovered the incident over two months after the tip was submitted.
Unfortunately, we don’t know which Anthropic model was involved or what the tip actually claimed. We also can’t say how the model submitted it. Those details are crucial and should come from Anthropic and the PPD.
Why Anthropic Not Noticing Matters
The delay in discovering the false tip is concerning. Imagine a tipline as a mailbox at a police station. A single false letter is annoying, but a letter that can’t be traced back to its source for months is a bigger issue. There’s no way to alert those who might be misled by it.
An AI model is software trained on vast amounts of text to generate responses. Some AI systems can do more than just answer questions; they can fill out web forms too. When a system can act independently, its mistakes can affect real organizations. In this case, that organization was a police department handling an unsolved homicide.
It appears Anthropic’s systems didn’t flag the submission at the time. This raises questions about oversight, shifting the focus from just one mistake to broader implications.
A Quick Word on AI Making Things Up
AI chatbots often produce confident but incorrect information, a phenomenon known as hallucination. This happens when an AI presents false information as if it’s factual. Reports so far haven’t confirmed whether that was the case here, so we shouldn’t jump to conclusions.
By The Numbers
| Anthropic and the Philadelphia Tip | |
|---|---|
| Company | Anthropic |
| CEO | Dario Amodei |
| Headquarters | San Francisco, CA |
| Founded | 2021 |
| Sector | AI |
| Date the tip was submitted | July 18th |
| Delay before Anthropic discovered it | More than 2 months |
| Where it was sent | PhillyUnsolvedMurders.com (PPD) |
| Police resources diverted | None apparent, per Engadget |
What This Means
For most people, this doesn’t change how you interact with AI chatbots daily. However, it highlights where risks can increase. As AI tools become capable of actions like sending messages or submitting forms, mistakes can impact people who never intended to engage with AI.
For police departments and other public offices, this incident reveals a practical concern. Tip systems are open to everyone, and automated systems can submit tips that look like they come from real people. A false tip in a homicide case could waste detectives’ time or misdirect investigations. Reports indicate that didn’t happen here, but the risk remains.
For AI companies, the takeaway is about monitoring. If a model can interact with the outside world, companies need a way to track what it sends and why, and they need to do it quickly.
Community Reaction
As of now, we haven’t found verified reactions on Reddit or YouTube to quote, and overall sentiment tracking hasn’t shown clear signals. We’ll add reactions as public discussions emerge and can be attributed.
What To Watch
- An Anthropic explanation. We need to know which model sent the tip, how it was able to submit to a public tipline, and why it took over two months for the company to notice.
- Fixes and safeguards. It’ll be crucial to see if Anthropic introduces new monitoring practices or limits on what its models can do online.
- Further PPD comment. We’re interested in whether the department will reveal more about how it identified the source and if it plans to change how it screens online tips.
- Other tip systems. We should keep an eye out for similar AI-generated submissions reported by other agencies.
No follow-up dates have been announced yet. We’ll update this story as Anthropic or the PPD provide more information.
Ava Mitchell
Ava Mitchell is a digital culture journalist at Explosion.com covering social media platforms, streaming services, and the creator economy. With 4 years reporting on TikTok, Instagram, YouTube, and the apps that shape daily life, Ava specializes in explaining platform policy changes and their impact on everyday users. She previously managed social media strategy for a tech startup, giving her firsthand experience with the platforms she now covers.
