Anthropic plans to change enterprise data retention policy: Report
Anthropic has been working with more than 100 customers, including Salesforce, on the planned system, which would allow enterprises to retain required data on their own cloud infrastructure.
Anthropic is preparing to change how it handles data from enterprise customers using its most advanced AI models, with a new system expected to give businesses more control over where retained data is stored, Reuters reported.
The Claude maker is expected to introduce the safety system later this year. Under the planned arrangement, enterprise customers would still have to retain data for 30 days, but they would be able to store that data on their own cloud infrastructure rather than Anthropic’s.
The system has been under development for several months. Anthropic has also been working with more than 100 customers, including Salesforce, as it develops the approach, the source said.
The planned change follows Anthropic’s decision in June to impose a 30-day retention requirement on enterprise traffic involving its more advanced Fable and Mythos models, as well as future frontier models. The company said the retention policy was intended to help protect against the potential use of its technology in cyberattacks.
The move also comes shortly after OpenAI introduced a safety system designed to detect potential misuse without retaining customer data. Anthropic’s proposed approach would therefore give businesses greater control over where their retained information is held while maintaining the 30-day retention requirement.
Anthropic also introduced text watermarking for future Claude models. The company says future Claude-generated text will contain a watermark that can help determine whether Claude was involved in producing it. “Watermarking does not impact the quality of Claude’s output,” the company said. “To a reader, a watermarked response is indistinguishable from an unwatermarked one, ” it added.