Major corporations, including Nvidia, Palantir Technologies, and Booz Allen Hamilton, are restricting the use of frontier AI models from OpenAI and Anthropic. This shift follows concerns regarding data privacy and the handling of proprietary information, particularly after Anthropic updated its policy in June to retain customer data for 30 days to detect potential abuse.
Nvidia has limited the use of Anthropic's Claude models to low-sensitivity tasks, such as open-source software development. For its most confidential business operations, the company now relies on its proprietary Nemotron models. Similarly, defense contractor Booz Allen Hamilton has implemented stricter internal controls and is seeking enhanced contractual protections for its data.
Other companies are adopting different strategies to mitigate risks. Telecom operator C Spire has signed agreements with both OpenAI and Anthropic to prohibit the use of its data for model training, though technical usage information is still collected. Pharmaceutical giant Novo Nordisk continues to use Claude for specific tasks while maintaining a strict ban on inputting proprietary data into the system.
Microsoft is positioning itself to capitalize on these security concerns by promoting isolated cloud environments. These setups run AI models on privately managed servers, ensuring that no user data is transmitted to external vendors. While this approach involves significant costs, with some deployments reaching millions of dollars, it meets the demand for secure processing of regulated and confidential information.
