What happened
The US Department of Defence confirmed to the BBC that it has stopped using Anthropic's AI products, fulfilling a directive issued by Secretary of Defence Pete Hegseth in February. Although the Pentagon designated Anthropic a national security supply-chain risk and set a deadline to stop usage by late August, multiple sources told the BBC that Claude models remained in active use for intelligence gathering and military operations as recently as last week. The removal of these tools from the Pentagon's primary data platform, Maven Smart System, marks the conclusion of a significant dispute over AI safety standards in military applications.
The US Department of Defence issued a statement to the BBC confirming that it has ceased the use of Anthropic products. This action follows a designation by Secretary of Defence Pete Hegseth in February, who labelled Anthropic a supply chain risk on national security grounds and announced a phase-out by late August.
Despite the official deadline, multiple people familiar with the matter, including former US defence officials and contractors, told the BBC that Anthropic's Claude model was still being used by the department as recently as last week. These sources indicated that the AI tools were utilized in research, analysis, intelligence gathering, and military operations against Iran.
The Pentagon had pressed Anthropic to remove safety from Claude to allow unfettered military access. When Anthropic refused, citing concerns about mass surveillance and autonomous weapons, Hegseth designated the company a national security supply-chain risk, a legal typically reserved for foreign entities.
Claude was deeply integrated into the Maven Smart System, a data platform operated by Palantir that serves as the Pentagon's primary tool for organizing intelligence. Experts noted that removing such deeply embedded AI systems is technically challenging and not a simple plug-and-play process.
During the period of the dispute, the Pentagon signed contracts with other major tech companies, including Google, xAI, and OpenAI. OpenAI tools have reportedly been more widely adopted in some military departments in recent months as alternatives to Anthropic's products.
Why it matters
This development resolves a high-profile conflict between a major AI developer and the US military over the deployment of safety in classified environments. It signals that the Pentagon is enforcing its supply chain risk designations, potentially setting a precedent for how other AI vendors are treated in government contracts. The shift also highlights the operational difficulty of removing deeply integrated AI systems from critical military infrastructure, as noted by experts who described the process as painful and complex. Furthermore, it underscores the ongoing tension between national security objectives and AI safety concerns, with the Pentagon now relying on alternative providers like OpenAI and Google for similar tasks.
The confirmation of the phase-out validates the Pentagon's authority to designate domestic AI companies as supply chain risks, a move that could influence future negotiations with other AI vendors regarding safety standards and government access.
The incident highlights the operational friction involved in integrating and subsequently removing AI systems from critical military infrastructure. The delay in fully removing Claude, despite the February order, suggests that such systems become deeply entrenched in workflows, making rapid divestment difficult.
This event underscores the ongoing debate over the role of safety in military AI applications. Anthropic's refusal to remove guardrails, based on concerns about autonomous weapons and surveillance, contrasts with the Pentagon's desire for unrestricted access, reflecting broader tensions in the AI industry.
The shift to alternative providers like OpenAI and Google may accelerate the adoption of these tools in military contexts, potentially altering the competitive landscape for AI in government contracts and influencing the development priorities of these companies.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
crm_get_transaction(id='4092').Why can ethical evaluation not be reduced to one model score?
What to watch next
Monitor the legal proceedings between Anthropic and the Pentagon regarding the supply chain risk designation. Observe how the Pentagon integrates alternative AI tools from OpenAI and Google into its existing workflows. Watch for any further statements from Anthropic regarding its stance on military AI use and safety . Track the impact of this decision on other government agencies that may have been using Anthropic's tools for classified work.
Legal outcomes of the dispute between Anthropic and the Pentagon, which may set precedents for how AI safety concerns are handled in government contracts.
The extent to which OpenAI and Google tools are integrated into Pentagon workflows, and whether these alternatives face similar scrutiny regarding safety and operational use.
Statements from Anthropic and other AI companies regarding their policies on military AI use and the balance between national security needs and safety .
Potential impacts on other government agencies that may have been using Anthropic's tools, and whether they will also phase out these products in response to the supply chain risk designation.