Does Claude Share What You Tell Cowork With Claude.ai?

TL;DR. Yes. As of August 25, 2026, Claude.ai chat and Claude Cowork share the same memory — so what Claude learns about you in one app is available in the other. Memory is on by default for Free, Pro, Max, and Team plans. You can review, pause, edit, or delete it in Settings → Memory. Sensitive topics (health, politics, religion) are excluded by default. There's no way to keep Chat and Cowork memories separate without using different accounts.

---

On August 25, 2026, Anthropic merged the memory system behind Claude.ai and Claude Cowork into one. A preference you mentioned in a chat last week is now available in your Cowork sessions. Context from a Cowork task will carry into your next chat. For many professionals, that's genuinely useful — you no longer have to re-brief Claude on your role, clients, or workflow every time you switch contexts. But it also means two apps that used to hold separate records of your conversations now share a single memory pool, and it's on by default.

Here's what actually changed, what gets stored, and what you can do about it.

What the memory merge actually means

Before this update, Claude.ai's chat memory and Cowork's session context were separate. You might have told Claude in chat that you're a paralegal who prefers plain-English summaries — but when you opened Cowork to run a research task, it didn't carry that context. You had to repeat yourself, or build project-specific instructions from scratch in each Cowork task.

That's what's changing. When you run a Cowork task in the cloud, Claude now pulls from the same memory it maintains in chat — your working preferences, recurring project context, communication style, whatever it's stored about you. And what Claude learns in a Cowork session can update that shared memory, making it available the next time you open chat.

One important limit: this only applies to cloud Cowork sessions. If you're running Cowork locally on your computer — using the desktop app with local file access or computer control — that session doesn't share memory. Only cloud-based Cowork tasks participate in the shared memory pool.

What gets stored — and what doesn't

Memory now works as a set of individual, categorized entries rather than a single daily summary. Claude stores things like your job role, active projects, communication preferences, recurring workflows, and professional context you've mentioned in conversation.

There's a meaningful default exclusion: sensitive personal information is not stored unless you opt in. By default, Claude won't write memory entries containing your health data, race, ethnicity, religious beliefs, political views, or gender identity. If you prefer Claude to remember that context — say, your dietary restrictions for a nutrition-related project — you can enable it under Settings → Memory → "Include sensitive topics in memory."

This exclusion reflects Anthropic's stated position on sensitive data, and it's worth knowing about if you've ever discussed health situations, personal beliefs, or anything adjacent to those categories with Claude.

Four things you can do right now

1. Review what's stored. Claude's memory entries are now organized by topic in Settings → Memory. It's worth a look before your next Cowork session — you may be surprised what's there, or what's missing.

2. Edit specific entries. You don't have to wipe everything to clean up outdated context. You can edit or delete individual memory entries while keeping the rest intact. If Claude stored something inaccurate about your role or a project that's since ended, you can fix it directly.

3. Pause memory. Pausing keeps your existing memory entries in place but tells Claude not to use them or create new ones during your sessions. It's a good option if you want to test a workflow without memory influencing the results, or if you're working on something you'd prefer Claude to approach fresh.

4. Delete memory entirely. This removes all stored entries. Claude will still function normally; it just won't have any accumulated context about you. Future conversations will start from scratch.

The one thing you can't do

There's no setting to maintain separate memory pools for chat and Cowork. You can't tell Anthropic "remember my work style in chat but keep Cowork clean." The systems are merged, and the only way to keep them separate is to use different Anthropic accounts — one for chat, one for Cowork. For most professionals, that's not practical. The realistic choice is between full shared memory, paused memory, or no memory.

What this means for your Cowork workflows

For most knowledge workers, the memory merge will make Cowork more useful out of the box. If you've already trained Claude to understand your role, clients, or preferred output format through regular chat, that context now carries into your Cowork tasks without any setup.

The catch is privacy review. Claude's memory is now a shared record that spans two apps — and if your Cowork tasks touch client data, project details, or anything you'd prefer to keep scoped to that session, it's worth checking what ends up stored. Memory entries are pulled from your conversations, not just things you explicitly tell Claude to remember, so context you mentioned in passing may be in there.

If you use Cowork with clients: consider whether any client-specific context from Cowork tasks should be visible to your Claude chat sessions (and vice versa). If you'd rather keep client work and personal chat separate, the two-account approach or periodic memory clearing is your cleaner option.

If you're on a Team or Enterprise plan: check with your admin. Enterprise accounts have organization-level memory controls, and your admin may have already configured what Claude remembers across accounts. Individual memory settings in these plans may be overridden by org policy.

On the privacy policy update

Separately from the memory feature, Anthropic updated its Consumer Terms and Privacy Policy on the same day. The update gives you a choice about whether Anthropic uses your conversation data — including memory-generating conversations — to improve Claude's models. You can opt out of this data contribution in your account's privacy settings. Memory and training-data use are configured independently: you can have memory on while opting out of model training, or have memory off while allowing your conversations to contribute to training.

The distinction matters for professionals in regulated industries. Memory stores context you can review and delete. Training data contribution is a different setting, and opting out doesn't erase past contributions — it stops future ones.

---

Sources

• Claude Cowork finally remembers what you told the app in chat — TechCrunch
• Anthropic's Claude and Cowork will share memories about you now — unless you opt out — ZDNet
• Updates to Consumer Terms and Privacy Policy — Anthropic
• Use Claude's chat search and memory to build on previous context — Anthropic Help Center
• Claude's memory now works across both chats and Cowork sessions — Engadget
• Claude and Cowork now share what they know about you — The Register