Anthropic announced a significant update to Claude’s memory system on August 25, 2026. The AI assistant’s memory now works seamlessly across both Claude Chat and Claude Cowork, the company’s cloud-based agent for multistep tasks. Users can also see, edit, and delete everything Claude remembers, organized by topic.
The update addresses a persistent friction point: having to re-explain context every time you switch between chatting with Claude and handing it a task in Cowork. That gap is now closed.
One Memory Across Claude Chat and Cowork
Cowork, Anthropic’s computer-use agent released earlier in 2026, now shares the same memory system as Claude’s chat interface. When Cowork runs a task in the cloud, it has access to everything Claude learned from your conversations, and vice versa.
The practical impact is straightforward. Ask Cowork to draft an update for your manager, and it already knows who that person is and how they prefer updates written. Brainstorm a conference agenda in chat, and when Cowork builds the budget and logistics document, it already has the headcount, city, and speakers. Explain once how your team defines its metrics, and every quarterly business review deck Cowork produces afterward uses them without rebriefing.
This unified memory eliminates the “two separate assistants” problem that previously made switching between chat and Cowork feel like starting over.
Memory Updates in Real Time
Claude now adds topics to memory as you chat, rather than summarizing conversations after they end. Mention that a project deadline moved to September, and your next conversation already knows without you having to say “remember this.”
This shift from post-conversation summaries to real-time topic capture means memories are more current and granular. Users can pause memory or reset it entirely at any time through the settings panel.
See and Edit Your Saved Memories
Everything Claude remembers is listed under Topics in Settings > Memory. Each entry is a short file organized by subject, and users can read, edit, or delete any of them.
The transparency goes both ways. If Claude stored your company’s old name, correcting it in one file fixes every future conversation. If you want to remove a topic entirely, deleting it takes effect immediately across both chat and Cowork.
This level of visibility sets Claude apart from some competitors that store memory in less accessible formats. Users always know exactly what the AI is working with.
Sensitive Topics: A Privacy-First Approach
By default, Claude does not store topics related to personal or sensitive subject matter, including health conditions, race, ethnicity, religious beliefs, politics, and gender identity. This is a deliberate design choice that prioritizes user safety over convenience.
However, Anthropic recognizes that some users want Claude to remember these details. A toggle in Memory settings labeled “include sensitive topics in memory” lets users opt in. When enabled, Claude saves sensitive topics going forward, and each save triggers a visible notice. Nothing from before the toggle was turned on gets saved retroactively.
Even with sensitive topics enabled, certain information is never stored. This includes Social Security numbers, government-issued IDs, criminal history, immigration status, and anything that violates Anthropic’s Acceptable Use Policy. Claude will inform users when it cannot update memory to include such information.
Getting Started
Memory is on by default for Free, Pro, and Max plans across web, desktop, and mobile. Saving sensitive topics in memory remains off by default. On iOS and Android, users need to update to the latest version of the Claude app to access the new features.
For Team and Enterprise plans, administrators control memory availability for their organization. Individual users on these plans have memory off until they explicitly turn it on.
To manage memory, visit Settings > Memory in the Claude interface.
How Claude’s Memory Compares to ChatGPT and Gemini
Claude’s updated memory system arrives in a competitive landscape where all major AI assistants are investing heavily in persistence:
- ChatGPT builds a persistent profile from conversations, typically storing 40 to 80 facts for a monthly user, and is rolling out a self-updating system.
- Gemini takes a different approach through Google’s “Personal Intelligence,” which draws on your entire Google account, including Gmail, Drive, and Calendar.
- Claude now organizes memory by topic with full user transparency and a privacy-first default that excludes sensitive categories.
According to a recent comparison of AI memory systems, each platform makes different tradeoffs between automation, privacy, and user control.
Claude’s approach stands out in several ways. The topic-based organization makes memory transparent and editable, rather than a black box of inferred facts. The privacy-first default, where sensitive categories are excluded unless explicitly opted in, is more conservative than competitors. And the unified memory across chat and Cowork means the AI’s context carries over whether you are having a conversation or delegating a complex task.
For users who value control and transparency over maximum automation, Claude’s memory system offers a distinct tradeoff: you see everything, you can change everything, and sensitive information stays out unless you say otherwise. As AI assistants increasingly build persistent memory layers into agentic workflows, the question of who controls that memory becomes just as important as whether it exists.
