US Tech Giants Restrict Use of Anthropic AI Models Amid Data Retention Concerns

Concerns over data retention and trade secret protection have prompted several major US technology companies to restrict their use of closed-source AI models, particularly those from Anthropic. According to a recent report, Nvidia has narrowed its use of Anthropic's models to non-sensitive business functions, while Palantir is demanding an 'irrevocable zero data retention' commitment from Anthropic before integrating its models.

The controversy centers around Anthropic's policy change in June, which allows the company to store user usage logs for 30 days. This has raised fears that core assets submitted to the model could be accessed or analyzed by Anthropic within that time frame.

Nvidia and other companies are taking a cautious approach, with Booz Allen Hamilton outright banning employees from using Anthropic's commercial models in classified cybersecurity work. Microsoft, on the other hand, is leveraging enterprise concerns about data security to pitch its isolated cloud environments and proprietary AI services.

The controversy has also sparked debate over the use of open-source models, which can be deployed locally and provide greater control over data and stronger privacy safeguards. With Anthropic's impending IPO looming, the company's attacks on Chinese open-source models are being seen as an attempt to prove the commercial value of its closed-source approach.