Google Rolls Out Computer Use for Mac Gemini
Google has rolled out a computer use feature for its Gemini macOS app, letting the AI assistant operate across local files, desktop programs, and the web to complete multi-step tasks. Available now in English for U.S. users, the system requires per-app consent and displays a live preview of every action. The launch precedes Apple's upcoming macOS 27.3 restrictions on full disk access for AI agents.


Short answer: Google has rolled out a computer use feature for its Gemini macOS app, letting the AI assistant operate across local files, desktop programs, and the web to complete multi-step tasks. Available now in English for U.S. users, the system requires per-app consent and displays a live preview of every action. The launch precedes Apple's upcoming macOS 27.3 restrictions on full disk access for AI agents.
Google Gemini Mac computer use launch
Google has added a computer use capability to its Gemini application for macOS, allowing the AI assistant to operate across local files, desktop programs, and the web to complete multi-step tasks. The company says the feature is rolling out now in English for users in the United States. With this update, Gemini can handle workflows that previously required manual switching between applications, such as gathering data from a spreadsheet, drafting a presentation based on that information, and then posting a listing to an online marketplace.
The system gives users control over which applications the assistant may access, and it provides a dedicated background window that displays a live preview of every action as it happens. This transparency layer lets people watch the agent navigate menus, type into forms, and manipulate files in real time without surrendering full oversight. Google suggests the feature could accelerate routine work like filling out repetitive forms, assembling slide decks, or preparing items for resale, though the company frames these as potential uses rather than guaranteed outcomes.
Apple Mac full disk access restrictions
The launch arrives just as Apple prepares to tighten restrictions on software that requests full disk access on the Mac. Those new rules, previewed in this week's macOS 27.3 beta, are a direct response to the potential risks posed by AI agents that can freely control a computer. By shipping computer use ahead of that clampdown, Google is positioning Gemini to operate under the current permission model while the platform holder finalizes stricter guardrails. The timing also highlights a growing tension between AI providers who want deep system integration and operating system vendors who must balance capability with security.
Desktop AI agents reach consumer software
For developers and teams building AI-powered products, the rollout signals that desktop agents are moving from research demos into shipping consumer software. The ability to chain actions across heterogeneous environments, local storage, native apps, and browser sessions, represents a significant step toward the kind of general-purpose automation that has long been promised. At the same time, the permission model Google has adopted, with per-app consent and a visible audit trail, may become a template for how other platforms manage agent access without exposing users to uncontrolled behavior.
Anyone who maintains macOS workflows or builds tooling that could benefit from scripted cross-application automation should evaluate the current Gemini release while the permission landscape remains relatively open. Apple's forthcoming restrictions will likely require new entitlements, additional user prompts, or sandboxing changes that could limit what agents can do without explicit approval. Early adopters who integrate the feature now will have a clearer migration path when those rules take effect, and they can provide feedback that shapes the eventual policy.
Enable Gemini computer use on macOS
If you are based in the United States and use the Gemini macOS app, you can enable the capability today and test it against your own repetitive tasks. Pay attention to how the live preview surfaces errors or unexpected navigation, because that visibility is the primary safety mechanism in the current design. For teams outside the initial rollout region, the priority should be designing workflows that assume a permission-first architecture, so they are ready when the feature expands or when similar agents arrive on other platforms.
Frequently asked questions
What is Google's new computer use feature for Gemini on macOS?
Google has added a computer use capability to its Gemini application for macOS that allows the AI assistant to operate across local files, desktop programs, and the web to complete multi-step tasks. The feature is rolling out now in English for users in the United States and can handle workflows like gathering spreadsheet data, drafting presentations, and posting listings to online marketplaces.
How does Google ensure user control and transparency with Gemini's computer use?
The system gives users control over which applications the assistant may access and provides a dedicated background window displaying a live preview of every action as it happens. This transparency layer lets people watch the agent navigate menus, type into forms, and manipulate files in real time without surrendering full oversight.
Why is Google releasing this feature ahead of Apple's macOS 27.3 restrictions?
Apple is preparing to tighten restrictions on software requesting full disk access on Mac, previewed in the macOS 27.3 beta, as a direct response to potential risks from AI agents that freely control computers. By shipping computer use ahead of that clampdown, Google positions Gemini to operate under the current permission model while Apple finalizes stricter guardrails.
What should early adopters and developers do with Gemini's computer use feature now?
Anyone maintaining macOS workflows or building tooling for cross-application automation should evaluate the current Gemini release while the permission landscape remains relatively open. Early adopters who integrate now will have a clearer migration path when Apple's new rules take effect, and they can provide feedback that shapes eventual policy.
How can U.S. users enable and safely test Gemini's computer use capability?
U.S.-based users of the Gemini macOS app can enable the capability today and test it against their own repetitive tasks. They should pay attention to how the live preview surfaces errors or unexpected navigation, because that visibility is the primary safety mechanism in the current design.
Source: The Verge
Related posts

Anthropic IPO Raises $81.3 Billion Amid Governance Questions
Anthropic completed its IPO on November 19, 2026, raising $81.3 billion-the largest ever, surpassing SpaceX. The AI lab, founded by ex-OpenAI staff prioritizing safety, adopted a Public Benefit Corporation structure with a Long-Term Benefit Trust holding controlling voting rights alongside founders. However, governance experts question the trust’s independence and AI expertise, noting its four trustees lack deep AI backgrounds and a recent trustee joined Anthropic as an executive, rais

Anthropic Goes Public on Nasdaq at $111.50 Per Share
Anthropic began trading on Nasdaq under ticker ANT on November 19, opening at $111.50 per share, above its $83 IPO price. The debut valued the AI company at roughly $1.5 trillion, the largest public offering on record. Despite reporting an $80.7 billion net loss for the first half of 2026, revenue reached $16.3 billion. Founders retain control through a dual-class structure. CEO Dario Amodei pledged to prioritize AI safety over profit pressure.

Teens Sue School Police After Doritos Bag Triggers AI Gun Alert
Four Black Baltimore County high school students were handcuffed and searched at gunpoint after an AI system mistook a Doritos bag for a gun. District officials canceled the alert within two minutes, but the untrained principal dispatched police anyway. Six patrol cars responded with weapons drawn. The teens have filed a federal lawsuit alleging unlawful search and detention, citing that over 99% of the district's AI alerts were false positives.