Anthropic plans to let customers store data on own servers

Anthropic plans to allow enterprise customers to retain data on their own cloud infrastructure rather than the company's servers, according to Bloomberg, a shift from a policy that has drawn pushback from businesses wary of sharing sensitive information with the AI startup.bloomberg

The Claude maker expects to roll out a new safety system later this year that will still require enterprise customers to retain data for 30 days but will give them the option to store it on their own cloud computing infrastructure, Bloomberg reported on Thursday, citing a person familiar with the matter. The system has been in development for months, and Anthropic began working on it before customers voiced criticism of the original policy, the person said.bloomberg

A Policy Under Pressure

Anthropic announced in June that it would require 30-day data retention for all prompts and outputs from its most powerful Mythos and Fable models to help detect and prevent cyberattacks. The company acknowledged the trade-off in a recent risk report, writing that the policy would "be unpopular with customers who have come to expect zero retention, and pose real risks to our business success (especially if competitors do not follow)."axios+1

The company has been coordinating with more than 100 customers in highly regulated industries to develop the new option, Bloomberg reported. Anthropic declined to comment.bloomberg

OpenAI Takes the Opposite Approach

The planned change comes one day after rival OpenAI announced it is testing a system called Private Safety Processing that aims to detect misuse patterns across related interactions without retaining customer data at all. OpenAI published a blog post on Tuesday outlining how the system sends only narrowly defined safety signals to the company, without exposing underlying prompts or responses.openai+1

"We're seeing with more capable frontier models that often risks are emerging not just by looking at one single prompt and response pair, but when you look over time at multiple interactions," Aleah Houze, OpenAI's head of product policy, told reporters in a briefing.axios

OpenAI said it plans a broader rollout and a technical white paper in September. The company is testing the system with early customers including Databricks and Microsoft ↘ 0.65%.thenextweb+1

Diverging Philosophies

The two approaches reflect a growing strategic divide between the leading AI companies. Anthropic maintains that retaining data — even on customer-controlled servers — remains essential for spotting sophisticated attacks that unfold across multiple requests. OpenAI contends it can achieve similar safety outcomes through cryptographic techniques that keep raw data out of its employees' reach.thenextweb+1

For enterprise buyers weighing which AI provider to trust with their most sensitive workloads, the distinction may prove decisive.