Palantir and Nvidia, along with other firms, are limiting how employees use Anthropic and OpenAI’s advanced AI models, citing concerns that proprietary customer or internal data could be retained or used to train future systems. The companies are seeking guarantees that AI providers do not keep data or learn from it before these models are used more broadly.
Quartz reports that Palantir and Nvidia push for assurances on data retention as they consider expanding access to frontier AI models. The Times of India reports that the restrictions extend to employee instructions to stop using the two providers’ models in certain contexts, and that the dispute is tied in part to Anthropic’s 30-day data retention policy introduced in June.
According to the Times of India, Palantir wants an “irrevocable zero” retention approach, while Nvidia routes sensitive work to its own models. Anthropic responds by pointing to its “Enterprise Frontier Safeguards,” which it says address the concerns. The outlets describe a broader effort by major firms to tighten terms with AI vendors rather than a change in the underlying AI systems themselves.
