Data Handling Concerns Reportedly Lead Palantir and Booz Allen to Limit Anthropic Model Usage

Several prominent government technology contractors, including Palantir Technologies and Booz Allen Hamilton, have moved to restrict how employees use Anthropic’s Claude AI models, according to a report from The Information. The restrictions stem from concerns about how the AI company handles data, according to the report.

Anthropic, one of the leading developers of large language models, counts several defence-adjacent firms among its enterprise customer base. The reported caution from contractors underscores a broader tension in the sector: while generative AI tools promise productivity gains in software development, analysis, and document processing, firms working with government clients often operate under strict data-handling and confidentiality obligations that make them unusually sensitive to supplier data practices.

The report also names Nvidia among the companies limiting use of Anthropic’s models. Nvidia, best known as the dominant supplier of the AI accelerators underpinning the industry’s buildout, has itself been expanding its own software and AI tooling ecosystem.

Palantir Technologies builds software platforms used by intelligence and defence customers, with its Gotham platform serving analysts in the United States, the United Kingdom, and internationally. The company’s shares traded at $170.50, down 0.07% from the prior close of $170.62, valuing the company at roughly $439.69 billion.

Booz Allen Hamilton, a major provider of consulting and technology services to US government agencies, has made AI adoption a visible part of its service offering in recent years. Its stock traded at $72.20, up 0.12% from the previous close of $72.11, giving the company a market capitalization of approximately $8.68 billion.

Neither the scale of the restrictions nor their precise mechanics were detailed in the report, which cited data-related concerns as the driver. Enterprise customers across regulated industries have increasingly scrutinized AI vendors’ data retention, training-use, and residency policies, with some negotiating bespoke contractual terms before allowing wide internal deployment.

For defence-sector companies, the calculus is sharper still. Contracts frequently involve sensitive government information, and vendors may be required to demonstrate that third-party tools cannot retain or repurpose client data. Reports of internal restrictions on a leading model provider illustrate how compliance departments, not just technology teams, are shaping AI adoption inside government services firms.

The episode also highlights competitive dynamics among AI labs, as enterprise buyers weigh providers on trust and governance alongside model performance.

What to watch

• Any confirmation or comment from Palantir, Booz Allen, Nvidia, or Anthropic on the reported restrictions
• Booz Allen’s upcoming earnings call, where AI-related contract work and spending trends may come up
• Palantir’s next quarterly results and any disclosures on its own AI platform adoption
• Broader enterprise AI procurement policies emerging across government services contractors

Source: original release