AWS Lets Bedrock AI Agents Keep 10 Years of Conversations — With Limits

Written By
eWEEK Staff
eWEEK Staff
Nov 17, 2026
3 minute read
Amazon Bedrock AgentCore Memory illustration showing conversation retention expanding from one year to 10 years with enterprise access controls

AWS extends AgentCore conversation retention to 10 years, giving enterprises more flexibility — and responsibility — over persistent AI memory. Image created using Google's Nano Banana 2

eWeek content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More

A decade is a long time to remember a conversation — even for an AI agent. AWS is making that possible with a new retention limit, though deciding who should have access to those records is another matter.

Amazon Web Services announced on November 17, 2026, that Amazon Bedrock AgentCore Memory can retain conversation events for up to 10 years, up from one year. The expansion could help enterprises preserve context for long-running support cases, recurring work, and audit reviews.

AWS also introduced automatic full-session summaries, helping agents recover context without reconstructing earlier conversations. Both updates simplify persistent memory management while increasing the amount of information organizations may need to protect.

How the 10-year retention setting works

The retention expansion applies to short-term memory events, which store conversations, rather than extracted long-term memory records such as preferences and summaries. Developers select a retention period for a memory resource; existing events keep their original expiration dates.

Advertisement

The maximum is optional and available in AWS Regions supporting AgentCore Memory. Under consumption-based pricing, AWS charges $0.10 per GB-month for short-term storage, $1 per GB for ingestion, and $0.20 per GB for retrieval. Ingestion and retrieval have minimum billable event sizes, so costs vary by usage.

Full-session summaries consolidate conversations into running records updated as new events arrive. They are created automatically for memory resources using the summary strategy, including previously summarized sessions, without separate configuration.

Persistent context does not guarantee better decisions. An October benchmark of enterprise AI agents found substantial task-completion gaps, although it did not test AgentCore Memory or its summaries.

AWS introduced fine-grained memory access controls in August 2026, enabling restrictions by user or tenant. Microsoft's Project Perception, introduced in limited public preview, takes a separate approach to managing agent identities, permissions, and changing security risks.

AWS also added automatic expiration controls for extracted long-term memory records in October. Those controls operate separately from the new conversation-event retention limit.

Advertisement

What eWeek Found: Tenfold retention, no separate summary permissions

Increasing maximum event retention from one year to 10 does not add separate authorization for full-session summaries. AWS's technical documentation confirms that these summaries share IAM actions and memory-resource permissions with other records.

Administrators cannot independently grant or deny summary access within the same memory resource. Applications can filter records by type, but filtering alone does not establish an authorization boundary.

An October Cloudflare AI data leak exposed customer information across isolated workloads despite authenticated access. The incident stemmed from shared inference infrastructure, not AgentCore Memory, but highlighted the need to test data separation beyond identity checks.

AWS has also documented memory poisoning, in which malicious stored content can influence subsequent agent interactions. Its sample write-validation pipeline supplements existing protections but is not a built-in, unified scoring service.

Advertisement

Enterprises considering decade-long retention must define access boundaries, validate stored information, and assess continuing storage costs. A support case may justify years of history; a routine agent interaction may not. Retention policies should follow workload requirements, not simply the longest duration AWS offers.

Want to learn more AI tips, tricks, and prompting techniques? Let us teach you How to Talk to AI for free! Try our six-minute course at The Neuron Academy, our practical learning platform designed to help professionals use AI more confidently at work.

Learn a few simple ways to write better prompts and get more useful results from AI, or browse our other AI course for free for seven days. eWeek readers get free 7-day access. Check out all the lessons here →

eWeek Logo

eWeek has the latest technology news and analysis, buying guides, and product reviews for IT professionals and technology buyers. The site's focus is on innovative solutions and covering in-depth technical content. eWeek stays on the cutting edge of technology news and IT trends through interviews and expert analysis. Gain insight from top innovators and thought leaders in the fields of IT, business, enterprise software, startups, and more.

Property of TechnologyAdvice. © 2026 TechnologyAdvice. All Rights Reserved

Advertiser Disclosure: Some of the products that appear on this site are from companies from which TechnologyAdvice receives compensation. This compensation may impact how and where products appear on this site including, for example, the order in which they appear. TechnologyAdvice does not include all companies or all types of products available in the marketplace.