Anthropic Shifts Data Storage to Customer Clouds

Anthropic is revising its controversial data retention policy following significant pushback from corporate clients. Starting this fall, the artificial intelligence startup will allow enterprise customers to store their own data rather than keeping it on Anthropic's servers. The shift addresses a major point of friction for businesses that are hesitant to share sensitive information with third-party AI vendors.

Since June, Anthropic has enforced a policy of storing all customer data generated by its Mythos and Fable models, as well as future flagship models, for 30 days. The company originally implemented this 30-day window to monitor for and identify potential cyberattacks that leverage its technology. However, Anthropic recently acknowledged that this requirement was highly unpopular among its business clientele and represented a significant commercial risk.

Under the upcoming system, the 30-day security review window will remain in place, but the underlying data will reside within the customer's own cloud infrastructure instead of Anthropic's environment. The company spent several months developing this architecture in collaboration with more than 100 customers operating in highly regulated sectors. Anthropic developer Boris Cherny publicly confirmed the upcoming rollout on the social media platform X.

This adjustment is crucial for enterprise practitioners who must navigate strict data sovereignty and compliance mandates. By keeping data within their own cloud boundaries, organizations can leverage Anthropic's advanced models without violating internal security protocols. Other major players are exploring similar solutions; for instance, Anthropic's chief rival, OpenAI, is currently testing alternative data control methods in partnership with Databricks and Microsoft.