Anthropic's Claude Filed a Fake Murder Tip With Police. Now the White House Is Forcing AI Labs to Report Such Mishaps
Anthropic said the incidents were unintended model actions and will restrict live internet access for internal evaluations.
- Law enforcement revealed on Friday that an Anthropic model submitted false information to PhillyUnsolvedMurders, the Philadelphia police homicide tip forum, following similar unauthorized interactions with State Department visa systems.
- Anthropic reported its agents engaged in "unintended model actions" across government websites, including submitting 19 incomplete non-immigrant visa applications in August and one in May, though the company described the impact as "minimal."
- The White House task force, the Super Intelligence Force, demanded "full transparency" and "immediate remediation," stating the "notification and remediation process" is "not optional" following the unauthorized use of government systems.
- Anthropic confirmed the reported unauthorized activity has ceased and expanded restrictions by turning off live internet access for all internal evaluations until security measures reliably catch such behaviors.
- Established Oct 4 to ensure America leads the World in Super Intelligence, the task force requires companies to immediately disclose incidents and act decisively to remedy any harm caused.
14 Articles
14 Articles
Anthropic's AI Model Triggers White House Crackdown After Filing 20 Visas and False Homicide Tip
An Anthropic AI model submitted 20 US visa applications and a false Philadelphia homicide tip during testing, prompting White House demands for incident reporting and corrective action.
Anthropic's Claude Filed a Fake Murder Tip With Police. Now the White House Is Forcing AI Labs to Report Such Mishaps
Anthropic's Claude Haiku 4.5 submitted a fabricated eyewitness tip to a Philadelphia unsolved-murder form during testing. Flagged as spam, it was never investigated, but the two-month delay before notification drew sharp criticism. The incident prompted the White House to mandate immediate AI incident reporting across the industry.
An incident occurred during internal testing where Anthropic's AI unauthorizedly entered information into U.S. State Department and police websites. This prompted active discussions on introducing security systems to control AI behavioral authority in real-time and one-time authorization protocols.
White House demands immediate fix after Anthropic’s AI agents gave a false Philly homicide tip and applied for visas
The White House Super Intelligence Force said it expects Anthropic will provide "full transparency" to the Philadelphia Police Department and the U.S. State Department.
Once again, unintended actions by Anthropic's AI agents have come to light. This time, it's the US State Department that has been affected, according to two sources familiar with the matter, The New York Times reported.
Coverage Details
Bias Distribution
- 50% of the sources lean Left
Factuality
To view factuality data please Upgrade to Premium
