Skip to content
Submit incident
AII-1001-58589
AllegedThe record is a lawsuit, complaint or accusation that has not been decided.

Pentagon used Anthropic's Claude to assist raid capturing Venezuela's Maduro

Reportedly, the U.S. Pentagon used Anthropic's Claude AI model through Palantir Technologies to assist in planning and executing a raid on Caracas, Venezuela to capture President Nicolás Maduro. The operation killed between 83 and 100 people and caused infrastructure damage, sparking a contractual dispute with Anthropic over the use of the system in lethal military operations that violated its stated acceptable use policy.

January 2026 · Pentagon · Safety and wellbeing
January 2026
AII-1001-58589View source ↗
Quick facts
Category
Safety and wellbeing
Source
wsj.com
Date
January 2026
Drafted by
Raidu Incident Detection

What happened

The Pentagon used Anthropic's Claude AI model, integrated via Palantir Technologies, to assist in planning and executing a U.S. special operations raid on Caracas, Venezuela to capture President Nicolás Maduro and his wife, Cilia Flores.

The raid resulted in 83 to 100 people killed, including security personnel and Cuban nationals, significant damage to infrastructure including localised power outages, a declared national emergency, and the removal of a head of state.

Reported impact

Between 83 and 100 people were killed in the operation, including Venezuelan security personnel and dozens of Cuban nationals.

The raid caused serious damage to the capital's infrastructure and triggered a state of national emergency in Venezuela.

Index analysis

Anthropic's Acceptable Use Policy explicitly prohibits using Claude for violence, weapons development, or mass surveillance, but the model was provided to the Pentagon through Palantir, creating a 'transparency fog' where Anthropic was allegedly unaware of the lethal mission.

The incident reflects tension between the Trump administration's 'AI-first warfighting' strategy, which prioritises speed and tactical dominance, and Anthropic's positioning as a 'safety-first' company.

Governance controls

Secretary of War Pete Hegseth threatened to use Defense Production Act wartime powers to force Anthropic to remove safety guardrails, treating AI safety protocols as a 'supply chain risk'.

The incident triggered a contractual standoff between the Pentagon and Anthropic over the use of the system in military operations outside the company's stated acceptable use policy.

Classification

Category
Safety and wellbeing
Industry
Government, security and defence
Harm status
Actual harm
Intent
Intentional
Timing
Post-deployment
Harm types
Physical death, Physical injury, Public interest, Human or fundamental rights
Harmed
General public, Government
Failure modes
Unsafe autonomous behaviour
AI systems
Claude

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

wsj.com
Pentagon used Anthropic's Claude to assist raid capturing Venezuela's Maduro