Anthropic will allow enterprise customers to keep their AI data on their own cloud infrastructure under an updated retention policy, according to a person familiar with the matter. The change marks a shift from the company’s previous stance, which required business clients to retain data for 30 days on Anthropic’s systems for models including Fable and Mythos.
The policy adjustment, expected later this year, follows months of development and coordination with over 100 customers, including Salesforce. Anthropic had announced in June that it would mandate 30-day retention of all enterprise customer traffic on its more advanced models and future frontier releases to mitigate risks of cyberattacks leveraging its technology.
The move aligns with growing enterprise demand for greater control over data storage and security. Rival OpenAI introduced a safety system on Wednesday that avoids retaining customer data entirely while still enabling misuse detection.
Anthropic’s updated approach provides clients with the option to deploy its AI models on their own cloud environments, reducing reliance on Anthropic’s infrastructure for data retention. The company has not disclosed additional details regarding the implementation timeline or technical requirements for the new policy.
