HHS plans no-bid enterprise AI deal with Anthropic

Advertisement

HHS plans to award Anthropic a sole-source contract to keep its workforce on the company’s Claude AI tools without an open competition.

The agency posted a notice of intent Oct. 19 on SAM.gov for a “limited continuity of services award.” The award would run until the department completes a competitive multiple-award blanket purchase agreement for enterprise frontier AI models. The notice did not list a contract value or projected award date.

HHS argued that only Anthropic can meet the requirement because Claude is the company’s proprietary model family and is already integrated into department infrastructure. Covered capabilities include chat, drafting and summarization, coding assistance, agentic features, connectors, and administrative controls such as single sign-on and audit export.

The notice also lays out data terms. HHS data is excluded from model training by default and stays in the continental U.S. The notice says Anthropic’s HIPAA business associate agreement posture is confirmed for surfaces eligible for protected health information. Anthropic must report security incidents within one hour.

The notice comes as Anthropic’s discounted General Services Administration OneGov deal is set to expire Oct. 31, FedScoop reported. HHS rolled out Claude across the department in December 2025 after Anthropic offered its services to agencies for $1 a year.

Jessica Tillipman, associate dean for government procurement law studies at George Washington University in Washington, D.C., told FedScoop the notice amounts to an admission that HHS depends on Claude. Ms. Tillipman said general chatbot services are a questionable basis for a sole-source award, and that the notice suggests the attitude “this is going to be a pain, so let’s just keep it going.”

HHS told employees in March to stop using Claude after President Donald Trump ordered federal agencies to phase out Anthropic’s technology during the company’s dispute with the Pentagon. The notice does not address that ban.

Advertisement

Next Up in Artificial Intelligence

Advertisement