Anthropic plans to give enterprise customers greater control over data generated by its advanced AI models, including the option to store information on private cloud infrastructure rather than Anthropic’s servers.
Under the revised policy, business clients will still be required to retain data for 30 days to mitigate cybersecurity risks, a requirement previously mandated for use of the company’s most powerful models, including Fable and Mythos, as well as future frontier models. The changes follow months of development and coordination with more than 100 customers, including Salesforce.
The update comes one day after rival OpenAI announced a safety system that avoids retaining customer data entirely while maintaining the ability to detect potential misuse. Anthropic’s policy adjustment reflects growing competition among AI providers to balance security obligations with client demands for data autonomy.
The company has not disclosed a specific timeline for implementing the new controls but indicated they will be rolled out later this year.
