Anthropic Plans 30-Day Claude Data Storage On Customers’ Own Cloud

Anthropic is preparing to give business customers greater control over data retained while using its Claude artificial intelligence models, in a planned shift to its enterprise data-retention approach. Under the proposed system, companies would still be required to retain certain data for 30 days, but they would have the option of keeping that information on their own cloud infrastructure instead of Anthropic’s servers.

The move comes as enterprise customers increasingly demand stronger control over sensitive AI data, particularly as companies use models for software development, financial analysis, healthcare and other business-critical workloads. Anthropic has reportedly been developing the new approach for months in consultation with more than 100 customers, including Salesforce, as AI companies face growing pressure to balance safety monitoring with corporate privacy requirements.

Anthropic Plans 30-Day Data Retention On Customer Clouds

Anthropic is expected to introduce a new safety system later this year that will allow enterprise customers to retain data for 30 days while storing it on their own cloud computing infrastructure.

The proposed change does not eliminate the 30-day retention requirement. Instead, it changes where the retained information can be stored. Businesses would have greater control over the infrastructure holding their data while continuing to meet Anthropic’s safety requirements.

The development represents a notable adjustment to Anthropic’s approach to enterprise data. The company has historically emphasized safety and security as important parts of its deployment strategy, but the retention of business data has become a point of concern for some large customers.

Why Businesses Want More Data Control

For companies deploying AI at scale, data governance can be as important as model performance. AI systems may process internal documents, source code, customer information, financial records and other commercially sensitive material.

Keeping retained information within a company’s existing cloud environment can make it easier for organizations to apply their own access controls, security policies and compliance processes.

Anthropic’s planned system could therefore make Claude more attractive to organizations that want to use advanced AI models while maintaining tighter control over where their information is stored.

Anthropic Has Been Working With Enterprise Customers

The planned change has reportedly been under development for several months, with Anthropic working with more than 100 customers to develop the system.

Salesforce is among the companies cited in reporting about the discussions. The involvement of large enterprise customers highlights the importance of data-retention policies as AI adoption moves deeper into corporate workflows.

The issue is particularly relevant for organizations that operate under strict regulatory or internal security requirements. Moving data storage to customer-controlled infrastructure can potentially reduce concerns about having sensitive information retained directly by an AI provider.

Anthropic’s existing commercial infrastructure already involves multiple cloud providers and geographic locations. Its privacy documentation says customer data for commercial products is stored in the United States by default, while processing can involve cloud providers in several regions.

Anthropic’s Broader Enterprise Push

The data-retention change comes alongside a broader expansion of Claude’s enterprise capabilities.

Anthropic has recently introduced features aimed at giving businesses more control over how Claude is deployed. These include the ability to run Claude Code sessions on a company’s own compute infrastructure and new enterprise security capabilities such as inference hooks for data-loss prevention.

These developments suggest that Anthropic is increasingly positioning Claude as infrastructure for enterprise AI rather than simply as a consumer chatbot.

Planned Change	What It Means For Businesses
30-day retention	Data would still need to be retained for the required period
Customer-controlled storage	Businesses could keep retained data on their own cloud infrastructure
New safety system	Designed to support Anthropic’s safety requirements
Expected rollout	Later in 2026
Customer involvement	More than 100 customers reportedly consulted

Safety And Privacy Are Driving The Debate

Anthropic’s retention policy is closely connected to the challenge of monitoring potentially harmful uses of advanced AI models.

As models become more capable, AI companies have argued that monitoring activity across interactions can help identify sophisticated misuse. However, such monitoring can require providers to retain information that enterprise customers may prefer not to leave on the provider’s infrastructure.

This creates a difficult balance. AI companies need enough information to enforce safety policies, while businesses want strong guarantees around confidentiality and data control.

Anthropic’s planned system appears intended to address both requirements by maintaining the 30-day retention period while shifting storage responsibility toward customer-controlled infrastructure.

Competition With OpenAI Is Intensifying

Anthropic’s policy shift comes at a particularly competitive moment for enterprise AI.

OpenAI recently previewed Private Safety Processing, a system designed to detect certain forms of AI misuse across multiple interactions while supporting deployments with zero data retention. OpenAI said the system is intended to preserve privacy while allowing safety monitoring to consider activity across sessions.

The development puts additional pressure on AI providers to offer enterprise customers flexible privacy and security options.

Businesses choosing between Claude, OpenAI and other advanced AI platforms increasingly have to consider more than model intelligence. Data retention, security architecture, compliance, deployment flexibility and control over customer information are becoming important parts of the purchasing decision.

Enterprise AI Is Moving Toward Customer-Controlled Infrastructure

The shift also reflects a broader change in how companies are deploying AI.

Rather than sending every piece of business information to a provider’s standard cloud environment, enterprises increasingly want options that allow them to maintain greater control over data and infrastructure.

This does not necessarily mean that businesses will run every AI model entirely on their own servers. Instead, hybrid approaches can allow companies to use powerful external models while keeping sensitive data within controlled environments.

Anthropic’s planned 30-day storage option fits into this broader trend.

Data Governance Could Become A Competitive Advantage

As AI becomes embedded in business operations, data governance is likely to become a major differentiator between AI providers.

A company may prefer one model over another not because it produces substantially better answers, but because its security architecture fits the organization’s regulatory requirements.

Anthropic’s decision to allow customer-controlled storage could therefore help address a major barrier to enterprise adoption: concerns about where business information resides after it has been processed by an AI system.

The Bigger Picture

Anthropic’s planned change illustrates the growing tension between AI safety and enterprise privacy. Advanced AI providers need mechanisms to identify misuse, but corporate customers increasingly expect control over sensitive information processed by those systems. Allowing businesses to retain data on their own cloud infrastructure could give Anthropic a way to address both concerns without completely abandoning its retention requirements.

The move also shows how enterprise AI competition is shifting beyond model benchmarks. As Claude and competing systems become embedded in business operations, customers are evaluating providers on security, compliance, infrastructure control and data-handling policies alongside performance.

Looking Ahead

Anthropic is expected to roll out the new safety system later in 2026, although the available reporting does not provide a specific launch date or detailed technical specifications. The success of the approach will depend on whether companies can maintain meaningful control over retained information while Anthropic can still meet its safety and abuse-monitoring objectives.

For enterprise AI customers, the development could mark a broader shift toward models that combine centralized AI capabilities with customer-controlled data infrastructure. As providers compete for corporate workloads, flexible retention and deployment options may become increasingly important factors in deciding which AI platforms businesses adopt.