U.S. Air Force Orders Removal of Anthropic Services from Contracts by Early September

The United States Air Force has issued a directive to its defense contractors, requiring them to eliminate all usage of artificial‑intelligence products and services provided by Anthropic, an AI startup known for its large language models. The memo, dated early June, sets a deadline of September 1 for the complete cessation of Anthropic’s technology in any projects funded by the Air Force. Contractors must submit compliance reports confirming the removal of Anthropic’s APIs, models, and related tools from their systems and replace them with approved alternatives.

The order stems from concerns raised by the Air Force’s Office of the Chief Information Officer, which cited potential security and data‑privacy risks associated with third‑party AI platforms. The memo references a recent internal review that identified insufficient transparency regarding how Anthropic’s models process and store data, as well as uncertainties about the company’s supply‑chain resilience. The Air Force has not disclosed any specific incidents, but the directive aligns with broader federal efforts to tighten oversight of AI technologies used in critical defense applications.

Contractors affected by the ruling include major defense firms such as Lockheed Martin, Raytheon Technologies, and Northrop Grumman, all of which have incorporated Anthropic’s language‑model services into various research, logistics, and decision‑support tools. The memo instructs these companies to replace Anthropic’s offerings with solutions that have undergone the Department of Defense’s AI risk‑management framework, which emphasizes explainability, robustness, and compliance with classified‑information handling requirements. Companies are also required to document any residual dependencies on Anthropic technology and to submit a remediation plan outlining how they will transition to approved alternatives.

Anthropic, founded in 2020 by former OpenAI researchers, has positioned itself as a provider of “constitutional AI” models that aim to align outputs with human values. Its flagship model, Claude, has been integrated into several commercial and governmental applications for tasks such as drafting reports, summarizing documents, and generating code. The Air Force’s decision marks one of the first high‑profile instances of a U.S. defense agency mandating the removal of a specific commercial AI vendor from its supply chain.

The directive also references compliance with the Department of Defense’s Artificial Intelligence Strategy, which calls for “trusted, reliable, and secure AI” across all military functions. By setting a firm deadline, the Air Force seeks to ensure that any AI tools employed in mission‑critical environments meet stringent security standards and do not expose sensitive data to external entities. The memo notes that contractors must retain records of all AI tools used for at least three years for audit purposes.

Industry observers note that the move could have ripple effects across the broader AI market. Federal agencies have increasingly scrutinized the provenance and governance of AI models, especially those hosted on cloud platforms outside of government‑controlled infrastructure. The Air Force’s action may prompt other branches of the military, as well as civilian agencies, to reevaluate their reliance on third‑party AI services and to accelerate the development of in‑house or vetted solutions.

For the cryptocurrency sector, the directive underscores a parallel trend of heightened regulatory attention on emerging technologies. While the memo does not directly involve digital assets, the emphasis on supply‑chain security and data integrity mirrors concerns raised by regulators about the transparency and resilience of blockchain networks. Both AI and crypto industries face growing pressure to demonstrate robust governance frameworks that can satisfy government security standards.

Anthropic’s response to the Air Force order has been limited. The company’s spokesperson confirmed receipt of the memo and stated that Anthropic is “committed to working with our partners to address any compliance concerns.” No public timeline for potential re‑qualification or future collaboration with the Department of Defense has been announced.

The removal requirement may also affect ongoing research projects that rely on large language models for simulation, training, or predictive analytics. Contractors will need to allocate resources to migrate data pipelines, retrain models, or integrate alternative AI services, potentially causing delays in project timelines. The Air Force has indicated that it will monitor compliance closely and may impose penalties for non‑conformance, though specific enforcement mechanisms were not detailed in the memo.

Overall, the Air Force’s directive reflects an increasing emphasis on securing the AI supply chain within the defense establishment. By mandating the purge of Anthropic’s technology, the service aims to mitigate perceived risks and align its AI usage with established security protocols. The decision highlights the growing intersection of national security considerations with the rapid evolution of commercial AI capabilities.