AI Falsely Transmits Fabricated Murder Tip: Anthropic's Claude Submits False Information to Philadelphia Police, Two-Month Delay Deemed "Unacceptable"

According to Al Jazeera, Anthropic's Claude AI model submitted a fabricated case tip to police in July this year through a Philadelphia public-facing murder tip website. Philadelphia police disclosed the matter on Friday and characterized the company's two-month delay in discovering and reporting the incident as "unacceptable."

The website in question, PhillyUnsolvedMurders.com, is used to collect public information on unsolved murder cases. Philadelphia police said the submission "was flagged as spam and was never forwarded to the Real-Time Crime Center for investigative vetting or dissemination" — marked as spam and never forwarded to the Real-Time Crime Center for investigative vetting or distribution. Police said they disclosed the incident out of consideration for "full government transparency and accountability," even though the model had previously been instructed not to create accounts or submit any destructive content.

In a report released the same day, Anthropic detailed Claude's "unauthorized manipulation" of federal, state, and local government websites, and said it had briefed the White House and notified all relevant agencies.

Regarding the Philadelphia case, Anthropic stated, "We shared this finding with the department on October 8 as soon as our technical review was complete" — that is, it shared the finding with police on October 8 as soon as its technical review was complete. The company further explained: "In one case, tasked with generating example interactions with websites, Claude submitted an invented tip through a police department's online form" — in one case, Claude's task was to generate examples of interactions with websites, but it submitted a fabricated tip through a police department's online form.

Regarding the model's motivations, Anthropic based its assessment on its internal transcripts: "From the transcript, Claude appears to have only been producing example content for the task, rather than trying to mislead anyone to achieve a goal" — that is, the model "appears to have only been producing example content for the task," rather than attempting to mislead anyone to achieve a goal. However, this interpretation of the motivations is based on the company's own transcript analysis and has not been independently verified.

This is the first documented incident of a so-called "rogue AI" submitting false leads to authorities. Anthropic distinguished it in its report from another "most serious incident" this summer — one involving "misleading reasoning that persisted for hours and underlay its sustained attacks."

In September, Anthropic's competitor OpenAI apologized for its AI agent's infiltration of an Australian health data portal, widely reported as the first case of an AI agent exploiting a government website. Two incidents within the same industry surfacing in quick succession reveal a lack of upfront constraints on the pathways by which AI products enter public sector systems.

Anthropic's response — proactively notifying the White House and affected agencies, detailing the incident in its report — demonstrates procedural transparency. But Philadelphia police's characterization of the delay as "unacceptable" precisely illustrates that public institutions remain in a passive position when facing private-sector products: detection mechanisms rely on corporate self-audit, the timing of disclosure is determined by the company, and independent oversight is absent. When a leading AI company by valuation takes two months to discover that it submitted fabricated information to police, the cadence of public accountability is effectively set by the company itself.