DoD Use of Anthropic's Claude
via Palantir During a
Supply-Chain Restriction
A formal public-interest oversight report documenting the critical discrepancy between the Pentagon's March 3, 2026 blacklisting of Anthropic and the continued operational use of Claude through Palantir's Maven Smart Systems in Operation Epic Fury.
The Human Cost of the Transparency Gap
The Committee has identified a profound discrepancy between the Department's March 3rd blacklist of Anthropic and reports that Claude remains integrated into military systems targeting over 42 targets per hour. This "Transparency Gap" has directly preceded the humanitarian, economic, and operational losses documented above.
Sequence of Events
Trump orders Anthropic blacklist
President Trump directs the U.S. government to stop working with Anthropic after an impasse over Claude's safety guardrails on autonomous weapons and domestic surveillance.
Minab Girls' School strike
A U.S. Tomahawk missile strikes an elementary girls' school in Minab, Iran, killing over 165 individuals — predominantly children. Preliminary findings link the strike to outdated intelligence.
Pentagon formally designates Anthropic a supply-chain risk
Defense Secretary Pete Hegseth signs the official supply-chain risk designation, ordering all DoD contractors to cease commercial activity with Anthropic, effective immediately.
Palantir confirms Claude still embedded in Maven
Reuters reports that Palantir's Maven Smart Systems — the Pentagon's flagship AI targeting platform — uses multiple workflows built on Anthropic's Claude. Palantir may require months to replace it.
Anthropic files federal lawsuits
Anthropic files two federal lawsuits against the Trump administration, alleging the Pentagon illegally retaliated against the company for maintaining AI safety restrictions.
KC-135 tanker crashes in western Iraq
A U.S. KC-135 Stratotanker supporting Operation Epic Fury crashes in western Iraq, killing all six crew members. Total U.S. service member deaths rise to 13.
Kharg Island strikes trigger oil price surge
U.S. forces execute large-scale precision strikes on Kharg Island, Iran's primary oil export hub. Brent crude surges above $100/barrel; New Zealand petrol prices rise ~$0.50/L.
Palantir CEO confirms continued Claude use
Alex Karp publicly confirms Palantir continues to use Anthropic's Claude in its tools despite the Pentagon blacklist, deepening the transparency gap.
Formal Oversight Request
Executive Summary
Pursuant to the oversight responsibilities of the House Armed Services Committee, the Secretary of Defense is requested to provide the materials and participate in the classified briefing described herein concerning the Department of Defense's contemporaneous designation of Anthropic as a "supply chain risk" and the ongoing operational use of the Claude model through Palantir in Operation Epic Fury.
Background & The Transparency Gap
In early March 2026, the Department of Defense officially designated Anthropic as a supply chain risk to national security, effectively blacklisting the company from direct government contracts. This decision followed a dispute over Anthropic's safety guardrails, which restrict the military's use of its technology for autonomous weapons and domestic surveillance.
Despite this designation, Palantir Technologies continues to utilize Anthropic's Claude model within its Maven Smart Systems — the Pentagon's flagship AI-powered intelligence analysis and targeting platform. Palantir CEO Alex Karp publicly confirmed on March 12, 2026, that Claude remains in use. Reuters reported that replacing Claude within Maven may require months of engineering work.
Admiral Brad Cooper, leader of U.S. Central Command, acknowledged on March 12, 2026, that AI had become "a key tool in helping the U.S. choose targets in Iran," stating that "AI tools can turn processes used to take hours and sometimes even days into seconds." This confirms that AI-assisted targeting — using a blacklisted vendor's model — was actively occurring during Operation Epic Fury.
Humanitarian Impact: The Minab Tragedy
On February 28, 2026, a U.S. Tomahawk missile struck an elementary girls' school in Minab, Hormozgan Province, Iran. Over 165 individuals were killed, the majority of them children. A preliminary U.S. military probe reportedly found that the strike may have been caused by outdated intelligence — precisely the type of error that rigorous human oversight of AI-assisted targeting is designed to prevent.
UNESCO condemned the strike, noting the school was struck twice within 40 minutes. The Human Rights Activist News Agency reported that by March 9, 2026, a minimum of 1,245 Iranian civilians had been killed in the broader conflict.
Economic Volatility: The Kharg Island Strikes
U.S. forces executed large-scale precision strikes on Kharg Island — Iran's primary oil export hub, responsible for approximately 90% of Iran's crude oil exports — on March 13–14, 2026. International benchmark Brent crude futures surged to $104.98 per barrel, up from approximately $73 per barrel before the conflict began.
Allied nations have experienced significant downstream economic effects. New Zealand's Finance Minister confirmed that petrol prices had risen approximately $0.50 per litre and diesel approximately $0.72 per litre since the conflict began disrupting shipping through the Strait of Hormuz.
Operational Losses
The intense operational tempo of Operation Epic Fury has resulted in significant losses for U.S. forces. Thirteen American service members have been killed in the conflict to date, with approximately 140 wounded. On March 12, 2026, a KC-135 Stratotanker crashed in western Iraq while supporting operations against Iran, killing all six crew members. The operational pacing enabled by AI-assisted targeting raises direct questions about whether the speed of operations has outpaced the capacity for human verification.
Legal Posture
The Committee is particularly concerned whether the Department's administrative record contains a contemporaneous, reasoned explanation reconciling the supply-chain designation with continued authorized operational use through Palantir. Absent such documentation, the Department's actions appear "arbitrary and capricious" under 5 U.S.C. §706(2)(A). Anthropic has itself filed two federal lawsuits against the Trump administration, alleging illegal retaliation for maintaining AI safety restrictions.
Requested Deliverables
A signed, dated chronology listing all supply-chain risk designations, waivers, Palantir contract amendments, and dates/times of Claude usage via Palantir since February 28, 2026. Secretary-level certification required.
All non-privileged internal memoranda, risk determinations, contracting officer determinations, and legal opinions used to designate Anthropic as a risk while simultaneously authorizing Palantir to provide access to Claude.
Diagrams illustrating where Claude inference occurs within Palantir's Maven system and the specific guardrails in place to ensure compliance with International Humanitarian Law (IHL) and civilian protection obligations.
Searchable API metadata (timestamps, account identifiers, model versions, classification flags) for March 1–16, 2026, specifically concerning operations in the Minab and Kharg Island sectors.
● All deliverables due within 3 calendar days. Secretary-level attestation required.
Key Locations
Interactive map of the principal locations referenced in this oversight report. Click any marker to view details.