A US official says the DOD stopped using Anthropic's tools; sources: Claude was in use as recently as last week, including in military operations against Iran
Technology reporter in San Francisco and — The US Department of Defence is no longer using Anthropic's AI tools …
Context & Ripple Effects
The break follows a dispute over military-use terms: Anthropic said proposed DOD language did not adequately prevent Claude’s use for mass domestic surveillance or fully autonomous weapons. That dispute mattered because sources had tied Claude to the Pentagon’s major air attack in Iran, while Anthropic later denied that it could manipulate a model once the military had deployed it.
The DOD decision also extends a broader federal separation from Anthropic. Treasury, State and the federal housing agency had already moved to end use of its products, with State identifying OpenAI as its replacement.
First-order effects
- The DOD loses access to Anthropic’s tools for its military workflows, while Anthropic loses a high-stakes government deployment for Claude.
- The department’s stop-use decision makes Anthropic’s military-use restrictions an immediate procurement constraint rather than solely a contractual dispute.
Second-order effects
- OpenAI gains a clearer opening across federal AI procurement after State’s announced switch, while other vendors face pressure to show they can meet agencies’ operational and policy requirements.
- Agencies using frontier models will need to weigh provider controls over surveillance and autonomous-weapons uses alongside model capability when selecting suppliers.
Third-order effects
- If federal buyers continue to favor providers whose deployment terms align with government missions, military AI procurement will increasingly sort labs by operational compatibility as well as technical performance.
- The dispute points toward dual-use AI governance being negotiated through access contracts: model providers retain leverage through use limits, while governments can shift demand to alternatives.
The trend: US government AI buying is becoming a test of whether frontier-model providers’ deployment rules are compatible with national-security use cases.