Anthropic changes data retention policy: What enterprise customers need to know

Anthropic on Tuesday announced changes to its data retention policy for business customers after strong feedback from enterprise users. The AI company claims it has been developing and fine-tuning a new system, spending hundreds of hours with clients to ensure companies have greater control over their data usage.

The new solution, called Enterprise Frontier Safeguards, will allow businesses to manage the review, storage and oversight of their data. The company said on its website that automated safety checks will run without requiring human review by its staff.

Why Anthropic changed its policy

Anthropic introduced a 30-day data retention requirement in June for traffic involving its advanced Mythos-class models. The move was made to assist in identifying misuse and defend against advanced and evolving attacks, the company said.

Anthropic also said that the information kept would not be used for any other reason than safety, such as training its AI models. However, privacy and data control were issues for enterprise customers.

What Enterprise Frontier Safeguards changes

The new solution will give companies more control of data. With automated safety monitoring, they can control the flow of information and its review. Monitoring can be done on a customer's own systems and data, according to Anthropic. This allows businesses to perform mandatory safety testing without giving Anthropic any access to the underlying data.

The company also announced that it won't charge customers for Enterprise Frontier Safeguards. The controls will support customers in directly deploying Anthropic's technology as well as deploying it via cloud providers.

What it means for Anthropic users

The change primarily affects enterprise customers utilizing Anthropic's advanced models. Instead of the 30-day data retention requirement, businesses will have more control over how data is handled.

The June policy is still outstanding, though, for non-enterprise customers who use Anthropic's Mythos-class models. Enterprise Frontier Safeguards will launch in phases and be generally available in fall 2026.

The transition is part of a strategy crucial to Anthropic's business, as enterprise adoption remains a key focus. It also comes after OpenAI's announcement of a new data retention system, which it says will help the rival AI firm attract customers facing tight privacy and security regulations.