Skip to main content
    Back to Blog
    12 min read
    Claude Cowork
    Claude
    Claude Enterprise
    Workplace Privacy
    AI Privacy

    Can My Employer See My Claude Cowork Activity? What Admins Get

    T

    Team Sonomos · Sonomos

    Team Sonomos writes about AI, privacy, and what happens to sensitive information at work.

    If you use Claude Cowork through a Claude Team or Enterprise plan your employer pays for, assume your employer can see what you do in it. On Enterprise that includes the full text: Anthropic's help center says Cowork sessions in Claude, Claude Desktop and Claude Mobile "are captured in the Compliance API," the tool an Enterprise organization uses to pull conversations for eDiscovery and investigations. Team plans have no Compliance API, but owners still get usage analytics and can stream Cowork activity events into their own security tools. On a personal Pro, Max or Ultra plan your employer has no admin console over the account, although a work computer can still record what happens on it.

    The detail that misleads people is local storage. When Cowork runs on your computer, Anthropic's article on Cowork for Team and Enterprise plans (updated November 16, 2026) says it "stores conversation history on users' computers," and that admins "cannot centrally manage or delete it." Read alone, that sounds private. The very next sentence reads: "Claude Enterprise admins can retrieve this content through the Compliance API." The history on your laptop is your copy. Anthropic keeps another, and your employer can ask for it.

    This guide sets out what Anthropic documents as of November 18, 2026, plan by plan, including the dates of the changes rolling out this quarter. Where the documentation is silent we say so rather than guess. For Claude chat outside Cowork, see Can my employer see my Claude chats?, and for every major AI tool side by side, see Can my employer see my AI chats?

    Where Cowork runs, and where the record ends up

    On Team and Enterprise plans a Cowork session runs in one of two places. Sessions in the cloud run on Anthropic's infrastructure, so work continues when the laptop is closed, and their "sessions and files are saved to the member's Claude account." Local sessions run on your own computer, with Claude's code in an isolated virtual machine, and keep their history there.

    For Enterprise organizations, the developer documentation on retrieving session transcripts explains how a local session still reaches the Compliance API: "Anthropic records each conversation server-side as its requests reach the Claude API; nothing is installed on the device." The transcript holds your prompts, Claude's responses, and each tool call with its result. It does not hold what never left the computer, so a file Claude never opened is not in it. Capture applies while the Compliance API is enabled and you are signed in with the Enterprise account.

    The same page sets the retention. Captured local session content is stored "for 6 years from capture by default," or for the organization's own retention period if it has set a finite one, and for 30 days in organizations with HIPAA readiness enabled. Sessions that ended before capture started for the organization are not listed, and sessions under zero data retention are excluded. The same endpoints cover Claude Code sessions run with an Enterprise login, which is worth knowing if you use both.

    What each plan gives your employer

    Anthropic splits these tools between the two business plans. Every row below comes from Anthropic's own help center or developer documentation.

    ToolWhat it gives the organizationPlan
    Compliance APIFull transcripts of local and cloud Cowork sessions, plus chats, files and projectsEnterprise; only the Primary Owner can switch it on
    Organization data exportConversation data and user data for the whole organizationTeam and Enterprise; Primary Owner only
    Usage analyticsCowork sessions per day and active users; on Enterprise, per-person daily counts through the Analytics APITeam and Enterprise
    OpenTelemetry exportA live stream of Cowork events to the company's own security tools, each tagged with your email addressTeam and Enterprise
    Inference hooksPrompts from supported Claude products sent to the company's security tool for a checkEnterprise, generally available since November 5, 2026
    EDR sensor in the Cowork VMThe company's endpoint security sensor reporting on activity inside the virtual machineTeam and Enterprise, beta on macOS and Windows

    The Compliance API is the one that matters most. Only the Primary Owner can enable it, but once it is on, the session endpoints need "no new key, scope, setting, or client update," and they return the transcript to whoever holds the organization's compliance key, usually a legal, security or compliance team.

    Inference hooks are newer. Anthropic's November 5, 2026 announcement describes prompt hooks that send prompts "to your security tool for a check" and pre-tool-call hooks that check tool calls before they run. In practice a company's existing data loss prevention policy can read a prompt before Claude answers it. The announcement does not list which Claude products support the hooks.

    Audit logs are the surface that does not cover Cowork. Anthropic's Cowork Enterprise admin guide says that "Claude Cowork activity isn't among" the audit log events and sends admins to the Compliance API for transcripts. The audit log article adds that chat titles and content are never exported in audit logs.

    What a Claude Team owner can see

    Team plans do not get the Compliance API or audit logs, which Anthropic describes as Enterprise features. That leaves a Team owner with less, but not with nothing.

    The first tool is analytics. Team plan Owners and Primary Owners can view usage analytics, including a Cowork page with sessions per day and daily, weekly and monthly active users. These are counts, not content.

    The second is OpenTelemetry. Cowork's OpenTelemetry export works on Team and Enterprise, and once an owner enters a collector address, "events begin flowing to your collector immediately." On current versions of Claude Desktop the events carry metadata by default: which tools ran, which folders you connected, whether you approved each action, token counts and cost. Your email address and connected folder paths are included "whatever the content capture setting." Prompt text, Claude's responses and tool inputs are added only if IT deploys a device policy called otlpContentCapture through its device management tool. Anthropic also warns that "some older versions of Claude Desktop send content when the policy isn't set."

    The third is the export. Anthropic's note to members, Who owns and manages the data of my team? (updated October 30, 2026), says the Primary Owner manages your work account and can request access to your data "through data exports, which may contain your conversations with Claude, uploaded files, and usage patterns." Anthropic does not say whether that export includes Cowork session transcripts. We could not confirm it either way, so do not assume they are left out.

    The unified Chat and Cowork switch, with dates

    Anthropic is merging chat and Cowork into one experience, and it turns on by default this quarter unless an owner has already chosen. In the unified version every conversation starts as a chat; when a task needs code or files, Claude continues it in a cloud sandbox Anthropic manages.

    • Enterprise. The Enterprise setup article says "Chat and Cowork unified" was off by default in existing organizations and "turns on by default on November 16, 2026" where no admin had set it. On December 1, 2026 it also switches on for eligible custom roles in organizations still on the default.
    • Team. The Team setup article says it "turns on by default on December 1, 2026, unless an owner keeps it off," and that the separate chat and Cowork experience is retired on February 1, 2027.

    For visibility, the change moves more work into the cloud and into the conversation record. The Enterprise article says "Chat history, sharing, export, the Compliance API, and your data retention setting all work from the conversation," and the Compliance API reference describes its coverage of unified chats as beta. Organizations with zero data retention, the HIPAA-ready configuration or customer-managed encryption keys cannot turn the unified experience on yet. One gap remains even after the switch: custom data retention does not cover "Cowork history stored on a member's computer."

    Does deleting a Cowork task remove it from your employer's view?

    Anthropic's Cowork getting-started guide says a deleted task leaves your task history immediately and is deleted from backend storage within 30 days, in line with its retention periods. That describes your history. It does not describe copies your employer already holds.

    On Enterprise, the Compliance API session endpoints are read-only, and local and remote sessions "cannot be deleted through them." A local session "has no server-side lifecycle status, and its visibility is governed by retention instead." The documentation does not say that deleting a task in the app removes the captured transcript, and it does not reconcile the 30-day figure with the six-year capture, so the cautious reading is that the compliance copy stays until its own retention runs out. Ordinary chats are spelled out: when a user deletes one, its content goes, but the Compliance API still lists the chat with a deletion date and an empty name.

    Anything already streamed through OpenTelemetry into a company's logging system, or exported by the Primary Owner, sits on the company's own infrastructure under the company's rules. Deleting the task in Claude does nothing to those copies. Our guide to AI chats as evidence explains why a company facing a dispute may be obliged to keep them.

    A personal Claude account on a work computer

    A personal Pro, Max or Ultra account is a contract between you and Anthropic. Your employer has no admin console over it and no Compliance API route into it. Three things still apply.

    The computer is the first. Endpoint monitoring and managed browsers see activity on a managed machine whatever account is signed in, and local Cowork history is stored on that machine. The second is that the employer may block you: Anthropic's Claude Desktop configuration article describes an Enterprise setting, "Limit Claude Desktop to allowed organizations (shadow IT)," that makes the app on a work computer refuse personal accounts. The third is the data. Pointing a personal Cowork session at a folder of company files sends company data to a provider your employer has no contract with, which is the shadow AI problem whether or not anyone ever reads the transcript.

    One naming note. Microsoft sells a separate product called Copilot Cowork inside Microsoft 365, and its transcripts are searchable with Microsoft Purview eDiscovery. If that is the Cowork you use at work, our Copilot guide is the one to read.

    Frequently asked questions

    Can IT admins see my chats inside Claude Cowork?

    On Claude Enterprise, yes. Cowork sessions are captured in the Compliance API, including sessions that ran on your computer, and the organization can retrieve the full transcript. On Claude Team, admins see usage analytics and any OpenTelemetry events they configure, which carry prompt text only if IT turns content capture on for your device.

    Is Claude Cowork history stored only on my computer?

    Not on Enterprise. A local session keeps its history on your computer, but Anthropic also records each conversation as its requests reach the Claude API and keeps that copy for six years by default where the Compliance API is enabled. On Team, Anthropic documents the local copy and the optional OpenTelemetry stream.

    Can my coworkers see what I prompt in Claude Cowork?

    Not through any feature Anthropic documents, unless you share the session and your organization allows sharing. The people who can see your prompts are the ones your company authorizes to use its admin and compliance tools.

    Does the unified Chat and Cowork experience change what my admin can see?

    It moves more agentic work into the cloud and into the conversation record. Anthropic says chat history, export, the Compliance API and retention settings all work from the unified conversation, with Compliance API coverage of those chats still in beta.

    How to use Cowork at work without surprises

    Start by checking which organization Claude Desktop is signed into, because the plan decides almost everything here. If it is your employer's Enterprise organization, write each Cowork prompt as though a compliance reviewer could read it years from now, since the default retention allows exactly that. If it is a Team organization, assume at least metadata reaches your security team and that prompt text can, depending on a device setting you cannot see. Keep personal tasks on a personal account and a personal computer, and choose connected folders with care, because every file Claude reads becomes part of the transcript. If you are the person setting this up, our Claude page covers the data side for teams.

    The other lever is what goes into a session in the first place. Locke, made by Sonomos, runs on your own computer and replaces names, account numbers and other sensitive values with placeholders before a prompt reaches the AI tools it screens, which include Claude on the web and the Claude desktop app, with coverage that depends on the app, platform and settings. It does not hide your Cowork use from your employer or make a work account private; it keeps those values out of what Claude receives. Locke is free for one person on x64 Windows and Linux, and you can download it here.

    Keep personal details out of what the AI receives

    Locke Free runs on your own computer. Before a prompt reaches the AI tools it screens, it replaces the names, account numbers, health details and other sensitive values it detects with placeholders, so the AI provider never receives them. Neither does Sonomos, which makes Locke and publishes this blog.

    It does not hide your AI use from your employer or make a work account private. Admin tools, retention and device monitoring work as this article describes, and monitoring software on a work computer can record what you type before Locke masks it. Detection can miss things, and coverage depends on the app and platform.

    Free for one person on x64 Windows and Linux. macOS is coming soon.

    This article is general information about privacy and AI, not legal advice, and reading it does not create a solicitor–client or attorney–client relationship. Regulatory requirements turn on your specific facts and jurisdiction and change over time. Consult qualified counsel before relying on anything here, and check the publication and last-reviewed dates above. We may revise posts, but we do not undertake to keep any post current as the law changes.

    Reference pages for this post