Anthropic plans new enterprise data controls for Claude customers

Anthropic is preparing a new system that would give enterprise customers greater control over how their data is retained when using the company’s most capable AI models.

The Claude maker expects to roll out the system later this year, according to Bloomberg, citing a person familiar with the matter.

Enterprise customers would still be required to retain data for 30 days, but they could have the option to store it on their own cloud infrastructure rather than Anthropic’s.

The system has been under development for months and Anthropic has been working with more than 100 customers in highly regulated industries to develop the option, according to the report.

Anthropic declined to comment.

The planned change follows a policy announced in June requiring customer data generated through its more powerful Mythos and Fable models, as well as future frontier models, to be retained for 30 days.

Anthropic introduced the requirement to help identify and prevent new forms of cyberattacks involving its models. The company said the retained data would not be used to train its AI systems.

Anthropic has acknowledged that the policy could create commercial challenges, particularly among customers accustomed to zero retention arrangements.

The move also comes as rival OpenAI develops its own approach to enterprise data privacy.

OpenAI said Wednesday that it had begun testing a system called private safety processing designed to conduct safety checks without retaining customer data.

The company is testing the system with customers including Databricks and Microsoft and plans to make it more broadly available in September.

Anthropic began developing its new system before customers started publicly criticizing the 30 day retention policy, according to Bloomberg.