Nvidia Palantir curb use of Anthropic models over data fears

Palantir Technologies, Nvidia and Booz Allen Hamilton are restricting or reconsidering their use of advanced Anthropic and OpenAI models amid concerns over customer data, intellectual property and data-retention policies. Palantir is seeking irrevocable zero-data-retention guarantees from Anthropic, while Nvidia limits Anthropic’s models to lower-risk tasks and Booz Allen bars their use for proprietary cybersecurity work. The concerns intensified after Anthropic authorized 30-day retention of usage logs for its Fable model and amid claims that OpenAI may have used customer data to help solve a mathematical problem. Anthropic and OpenAI say they do not train on enterprise data by default, but both collect anonymized or de-identified metadata, which continues to trouble corporate users. Microsoft is seeking to capitalize on the uncertainty by offering isolated cloud environments and its own AI products, while other companies are keeping sensitive work on air-gapped or in-house systems.