Hot AI News

Claude can now drive desktop apps for Team and Enterprise, if an owner flips the switch

Anthropic's computer use is now in beta on Team and Enterprise plans, off by default, with owner-only controls, app block lists and an MDM kill switch.

Claude can now drive desktop apps for Team and Enterprise, if an owner flips the switch
Image: Anthropic

Anthropic has put computer use in beta on Team and Enterprise plans, letting Claude see a user’s screen and work in the desktop apps that person allows. The release notes dated November 20 say it is rolling out to organizations gradually.

Off by default, and only owners can turn it on

According to the release notes, computer use works in the new Claude experience in Claude Desktop. It also works from Claude on the web and Claude Mobile when Computer access is switched on. Anthropic’s separate setup article for Team and Enterprise adds the administrative detail.

Claude can now drive desktop apps for Team and Enterprise, if an owner flips the switch
Image: Anthropic
  • The feature is off by default. Only an Owner or Primary Owner can enable it, under Organization settings, then Capabilities, after turning on the unified Chat and Cowork setting.
  • Claude Desktop must run on macOS 14 or later, or on Windows. Web and mobile chats need a computer with Claude Desktop open to connect to.
  • It does not work inside Cowork tasks or Claude Code, only in the new unified experience.
  • Organizations using the HIPAA-ready configuration, Zero Data Retention for Claude Code, customer-managed encryption keys, or FERPA and K-12 settings cannot enable it.

On Enterprise, built-in roles get access once the setting is on, while custom roles need one that includes the unified Chat and Cowork capability. Team plans have no custom roles, so the setting applies to everyone in the organization. The setup article also says changes can take about an hour, sometimes longer, to reach each user’s app.

Video: Anthropic
Advertisements

Block lists, a kill switch and one blunt warning

Owners can set an App access list, either blocking named apps or allowing only named ones. Users still approve each app themselves. The lists identify apps by macOS bundle ID or Windows file name, and Anthropic says nothing checks that a typed ID is real. They do not cover Claude in Chrome, the built-in browser, Claude Code, desktop extensions, websites, remote desktops or virtual machines.

Claude can now drive desktop apps for Team and Enterprise, if an owner flips the switch
Image: Anthropic

IT teams that use device management can also set a computerUseEnabled policy key to false. That requires Claude Desktop 5.1231.0 or later, and the app reads the key only at startup. Anthropic says the key does not cover Claude in Chrome, Claude Code in a terminal, local MCP servers or desktop extensions, and it is not a lock against someone who can change the computer’s settings.

On macOS the feature needs Accessibility and Screen and System Audio Recording permissions. A management profile can pre-grant Accessibility, but Anthropic says it cannot switch on Screen Recording. Windows needs no operating system permission.

Claude can now drive desktop apps for Team and Enterprise, if an owner flips the switch
Image: Anthropic

The setup article is blunt about the risk, stating there is no sandbox between Claude and a user’s applications. The folder limits and read-only setting that apply elsewhere do not apply to computer use. Screenshots are saved with the chat and may hold personal or confidential data. Anthropic adds that its automated action check can see the latest screenshot, and that a copy can be kept in the record of that check. Data handling falls under the Commercial Terms and Data Processing Addendum.

Gemini got there first on consumer Macs, but only in the US

The angle for rivals is timing and audience. Google began rolling out Gemini computer use on Mac on November 19, one day before Anthropic’s entry. According to our report on that release, it is for English-language personal Google accounts held by users 18 and over in the United States only. It requires a Google AI Plus, Pro or Ultra subscription, runs in a separate window with a live preview, and lets the user approve sensitive steps. The two moves target different buyers: Google is selling to paying individuals in one country, while Anthropic is aiming at managed workplaces, with admin policy as the main feature.

Anthropic is not new to the idea. Per its release notes, a computer use research preview reached Pro and Max users in Cowork and Claude Code on March 23, 2026. Its Cowork help article says Team and Enterprise users get computer use only in the new experience after an owner enables it. That article also describes how Claude behaves in practice: it tries connectors first, then the browser, and uses screen interaction last. Some apps are visible but not clickable, including browsers and many trading and crypto apps, and terminals and code editors allow clicks but not typing. Anthropic advises against using it for financial accounts, legal documents or medical information.

Advertisements

What the announcement does not establish

The release note gives no pricing, no seat or usage limits, no list of regions, and no date by which every organization will see the setting. Anthropic’s help pages say only that the rollout is gradual and that the toggle may not be visible yet. No accuracy or speed figures were published with the beta, so how reliably Claude completes multi-step desktop tasks in a real company remains unmeasured outside Anthropic’s own descriptions. Anthropic’s Cowork article itself notes that complex tasks may need a retry and that screen interaction is slower than using connectors.

For administrators, the practical takeaway is that nothing changes until an owner acts. Teams that want to test it should review the app block list and the screenshot handling terms first, and check whether their compliance setup is among those excluded from the beta.

Sign Up for the Daily AI Pulse

One email a day. All the stories that matter.

Alex CarterAI News & Big Tech Correspondent

Alex Carter writes for Aiholics, keeping readers updated on the fast-paced world of AI and Big Tech.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may also like