Anthropic announced that it will roll out a new safety system later in 2026, giving its enterprise customers greater control over the data generated when using the company’s advanced AI models. The revised data‑retention policy continues to require business clients to retain all AI interaction data for a minimum of 30 days, but now offers the option to store that retained data on the client’s own cloud‑computing infrastructure rather than on Anthropic’s servers. The changes have been under development for several months, with Anthropic coordinating the design and implementation with more than 100 enterprise customers, among them Salesforce, to ensure the solution meets varied operational requirements. Anthropic previously disclosed in June that a 30‑day retention rule would apply to traffic on its more powerful Fable and Mythos models, as well as on future frontier‑model offerings, a measure intended to guard against potential cyber‑attacks leveraging its technology. Bloomberg News first reported the development earlier on Thursday, and the Reuters report confirms the forthcoming rollout. In parallel, rival OpenAI announced its own safety system on Wednesday, which is designed to avoid retaining any customer data while still enabling the detection of potential misuse of its AI services. The contrasting approaches highlight a competitive focus on data‑privacy and security features within the enterprise AI market.