Nvidia, Palantir, Booz Allen Curb Anthropic Use Over Data Fears

Large technology companies — including Palantir Technologies, Nvidia, and defense consulting giant Booz Allen Hamilton — have begun restricting or demanding new guarantees before continuing to use Anthropic's and OpenAI's most advanced AI models, driven by rising concern that the labs could learn from customers' proprietary intellectual property, The Information reported.

Palantir has pressed Anthropic for irrevocable zero-data-retention guarantees before making its models available through Palantir's software. Nvidia has limited Anthropic's use to less sensitive tasks, relying instead on its own Nemotron models for internal work. Booz Allen has gone further, barring employees from using Anthropic's commercial model for proprietary cybersecurity work entirely — a restriction reinforced by the firm's own Cyber Weapon Index, published in early September, which found Anthropic's Claude Mythos model could autonomously execute full cyber kill chains, identifying, penetrating, and exploiting network vulnerabilities without human intervention at each step.

The pushback follows a June policy change to Anthropic's Fable model, which gave the company rights to retain usage logs for 30 days to guard against "complex and novel attacks" — a shift that triggered customer concern despite both labs maintaining they don't train on customer data by default unless companies opt in.

The timing compounds the unease: Anthropic CEO Dario Amodei recently called for AI companies to slow the pace of frontier development, warning capabilities are advancing faster than researchers can understand or control — a position quickly echoed by OpenAI's Sam Altman and Elon Musk.

Microsoft is capitalizing directly on the anxiety, pitching isolated cloud environments and its own AI offerings to draw customers away from both labs.

For defense contractors handling classified information, the underlying logic is stark: a model capable of autonomous offensive cyber operations needs to be owned and controlled, not rented from a company that could revoke access or change terms — reshaping who gets to build AI for the world's largest military and enterprise customers alike.