Concerns regarding data privacy and intellectual property protection are leading major corporations to restrict their use of advanced AI models. Companies are increasingly wary of how AI providers handle sensitive business information and metadata.
Anthropic updated its policies in June, allowing the retention of customer data for the Fable model to prevent misuse. This change has prompted some enterprises to seek alternatives, as they fear sensitive data could be compromised despite assurances that information is not used for training.
Telecom provider C Spire maintains agreements with OpenAI and Anthropic to prevent model training on its data, yet remains concerned about the collection of technical usage metadata. The company is seeking greater transparency regarding what information is gathered during AI interactions.
To mitigate risks, Northrop Grumman has opted to run open-source models on air-gapped servers. Meanwhile, Microsoft is promoting its secure, isolated cloud environments as a private alternative, though this approach involves higher costs for users.
Some organizations have implemented strict usage policies. Novo Nordisk prohibits the input of proprietary data into Claude, while Nvidia limits the use of Anthropic's Fable to non-sensitive tasks, relying on its own in-house AI solutions for critical operations.
The lack of nonrevocable zero data retention policies has already impacted business deals. A major U.S. utility company canceled plans to test the Fable model after failing to secure such a guarantee from Anthropic.
