US Government Blacklists Anthropic, Barring Agencies and Contractors from Its AI Tech

The blacklist stems from a formal determination by the Department of Defense’s Joint Artificial Intelligence Center (JAIC), which classified Anthropic’s Claude models as a “high‑risk supply chain component” after a security review uncovered undocumented data‑flow pathways that could expose classified information, FXLeaders reported. The assessment triggered an immediate prohibition on the use of any Anthropic‑derived APIs across all DoD agencies and their contractors, effectively removing Claude from the Pentagon’s emerging AI stack that had previously been piloted for natural‑language summarization and threat‑intelligence analysis.

The move was amplified by a directive from former President Donald Trump, who, according to Reuters, instructed federal agencies to cease all deployment of Anthropic’s technology “effective immediately.” The order cited concerns that the startup’s codebase could be leveraged by foreign adversaries, echoing the Pentagon’s earlier supply‑risk warning. While the Trump administration’s memo did not provide new technical details, it reinforced the JAIC’s stance and mandated compliance checks within 30 days, a timeline that agencies reportedly struggled to meet given the integration of Claude into several legacy systems.

Industry analysts cited by CNBC noted that the Pentagon’s pivot away from Anthropic could have broader ramifications for the company’s federal revenue pipeline, which had been projected to reach “hundreds of millions” in the next fiscal year. The outlet highlighted that Anthropic’s leadership had previously engaged with the Defense Innovation Unit to tailor Claude for secure environments, but the recent blacklist “places the startup in a lose‑lose situation” as it must now renegotiate contracts under stricter compliance frameworks or risk losing the entire defense segment. No alternative contracts have been publicly announced, and the company’s spokesperson declined to comment on the blacklist’s impact on its financial outlook.

In the wake of the ban, OpenAI secured a separate agreement with the Pentagon, as detailed by Tom’s Hardware, which described the deal as a “strategic partnership to provide a vetted, enterprise‑grade large language model” after Claude was removed from the approved list. The article noted that OpenAI’s GPT‑4o model, already cleared for classified use under a separate authority, would now fill the functional gap left by Anthropic, positioning OpenAI as the primary federal AI vendor. This shift underscores a broader trend of the U.S. government consolidating AI procurement around a limited set of providers deemed compliant with stringent security standards.

The blacklist also raises questions about the future of AI supply‑chain governance. The JAIC’s classification framework, which evaluates vendor code provenance, data residency, and vulnerability disclosure practices, is expected to become a de‑facto benchmark for other federal entities, according to the FXLeaders analysis. Companies seeking government contracts will likely need to adopt “zero‑trust” architectures and undergo continuous third‑party audits to avoid similar prohibitions. Anthropic’s experience serves as a cautionary tale that technical compliance, not just product performance, will dictate market access in the increasingly regulated AI landscape.