Nvidia, Palantir, Booz Allen Limit Anthropic Over Data Logs

Three of the most sensitive-data-heavy customers in enterprise AI have pulled back on Anthropic's models, and the trigger is a paragraph in a June policy update. The Information reported that Nvidia, Palantir and Booz Allen Hamilton are all restricting use of Anthropic's commercial model after the company reserved the right to retain usage logs for 30 days to guard against "complex and novel attacks."

Palantir has been pressing Anthropic for "irrevocable zero-data-retention guarantees before making its models available through Palantir's software." Nvidia "limits Anthropic's use to less sensitive tasks and relies on its own Nemotron models for internal work." Booz Allen has "barred employees from using Anthropic's commercial model for proprietary cybersecurity work."

Anthropic's public line is that it "does not train its models on information provided by enterprise customers by default unless they opt in," and it has announced a program letting eligible businesses store certain data on their own servers rather than on Anthropic's.

The pushback lands the same week Anthropic is reportedly in talks to bring Nvidia in as an anchor investor for its IPO, and days after Nvidia and Palantir shipped a sovereign AI stack whose Nemotron piece is pitched on "controlled deployments where sensitive data never leaves the customer's network."

The reporting is single-sourced so far, and Anthropic has not publicly named the customers.