Pentagon’s “All Lawful Use” Demand Causes Uproar

Close-up of a smartphone displaying the word 'ANTHROPIC' with a blurred portrait in the background

One Silicon Valley AI company just learned the hard way that refusing the Pentagon’s “all lawful use” terms can get you branded a national-security risk and cut off from federal business.

Story Snapshot

The Pentagon labeled Anthropic a “supply-chain risk,” threatening to end or restrict federal work tied to its Claude AI model.
Anthropic’s leadership says it set red lines against uses linked to mass surveillance or autonomous weapons, and it has vowed a legal fight.
President Trump ordered an end to Defense Department ties with Anthropic as OpenAI moved quickly to secure a competing Pentagon deal.
Claude was deployed through Palantir’s Maven Smart System in strikes involving Iran, intensifying the debate over battlefield AI.

Pentagon “Supply-Chain Risk” Label Raises the Stakes for AI Vendors

The Pentagon’s dispute with Anthropic escalated in early March 2026 after the Defense Department—rebranded by Secretary Pete Hegseth as the “Department of War”—designated Anthropic a supply-chain risk. That label can effectively freeze a firm out of sensitive government partnerships, even beyond a single contract. It tied the immediate jeopardy to a classified deal worth about $200 million, with broader implications for future federal procurement.

Defense officials argued that Anthropic’s proposed restrictions on how Claude could be used created operational and national-security problems, because the military wants tools deployable “for all lawful use,” including at high classification levels. Anthropic countered that its product limits are meant to prevent predictable abuses—especially surveillance applications and weaponization paths that could lower human accountability. The clash shows how quickly procurement leverage turns philosophical disagreements into existential business risk.

Trump’s “Patriotic Provider” Signal Meets Silicon Valley’s Safety Culture

President Trump’s decision to end Defense Department ties with Anthropic hardened the message that contractors working with the federal government must meet mission requirements rather than impose outside policy vetoes. In conservative terms, that reflects a familiar principle: elected leadership sets defense policy, not private executives or employee factions. At the same time, the story highlights a real tension—advanced AI tools can expand state power, and Americans have long demanded constitutional guardrails when surveillance capabilities grow.

Anthropic’s CEO, Dario Amodei, publicly indicated the company would fight the designation, while also apologizing after a memo criticizing rivals and the administration became public. It also points to personal rivalries between tech leaders shaping public messaging, including sharp exchanges between Amodei and OpenAI CEO Sam Altman. The result is a muddled public debate where genuine questions about oversight get mixed with corporate positioning and reputation management.

Battlefield AI Use Accelerates as Oversight Questions Lag

Multiple reports describe U.S. forces using Claude through Palantir’s Maven Smart System in operations related to Iran, with AI compressing parts of the targeting cycle that once took much longer. That kind of acceleration is exactly why the Pentagon wants fewer restrictions: speed can be decisive against capable adversaries. The governance question is whether faster decision loops will also compress scrutiny, documentation, and accountability—especially if systems are used across multiple theaters and classification environments.

Some details remain contested. Axios reporting described a disputed incident involving a raid connected to Nicolás Maduro that Anthropic allegedly learned about and objected to, while the Pentagon disputed key elements. That uncertainty matters because it affects the core factual question of visibility and consent: did Anthropic know how its model was used, and did it have a contractual basis to limit that use? Without clearer documentation, outside observers are left parsing competing claims rather than settled evidence.

OpenAI’s Pentagon Deal—and the Unverified “Worker Support” Narrative

OpenAI moved to secure a competing Pentagon arrangement after the administration’s decision, reflecting a broader shift that began when OpenAI removed military-use bans in 2024. That strategic change has been widely interpreted as prioritizing productization and large contracts over earlier, stricter posture. Meanwhile, the user’s prompt references claims that OpenAI workers supported Anthropic and that Anthropic could lose $5 billion. Those points are not confirmed with specific, on-the-record evidence, and even the “$5 billion” figure appears mismatched to near-term contract exposure.

Workers at OpenAI show support for Anthropic as the company says it could lose $5 billion in its feud with the Pentagon https://t.co/CaB4U8FR2E #news

— Business News (@15MinuteNewsBus) March 10, 2026

What is clear is the precedent the Pentagon is trying to set: “all lawful use” as a baseline expectation for AI contractors. For Americans wary of government overreach, that standard should trigger two parallel demands—military readiness against China and other adversaries, and tighter civilian oversight so AI doesn’t become a backdoor for unconstitutional surveillance or unaccountable lethal decision-making. The debate is moving fast, and procurement pressure is forcing every major lab to pick a side.