Skip to main content
AI News & Updates

GitHub Copilot Browser Use Enters Preview: Test Signed-In Sessions and Permissions Read-Only First

A connected browser does not authorize form submission on the user's behalf

ENHE AI5 min2 views

GitHub released Browser Use in public preview on November 17, 2026 for Copilot CLI, the Copilot app, VS Code, and JetBrains, paired with the GitHub Copilot Chrome extension. This guide starts with a read-only acceptance path. Use a test account and a fixed page, confirm that the extension reports Connected, record the tab scope and site permissions, and compare the elements and screenshots the agent uses with the expected page. Then verify that sign-in, form submission, deletion, and external communication stop for explicit approval. Expand only after the read-only flow is repeatable, permissions can be revoked, and failures do not cause unintended actions. The article does not claim a hands-on browser-use test or permission behavior beyond GitHub's documentation.

More topic tags (9)
AI NewsAuto PublishingSEOGitHub Copilot Browser UseChrome扩展Automation权限验收登录态只读测试
GitHub Copilot Browser Use Enters Preview: Test Signed-In Sessions and Permissions Read-Only First

Key takeaways

Start with a test account and a fixed read-only page.
Record tab scope, site permissions, screenshots, and element evidence.
Require approval before sign-in, submission, deletion, or external communication.

By ENHE AI · 2026-11-21 · AI-assisted drafting; pre-publication quality checks completed.

Step one: build a repeatable read-only baseline

Choose a test account, a fixed page without sensitive data, and three read-only actions: open the page, read one state, and locate a button without clicking it. Record the human expectation before Browser Use runs.

Pass only when repeated runs stay within the same tab scope, return the same facts, and identify whether evidence came from page elements or screenshots. “The agent saw it” is insufficient.

Step two: verify connection and permissions

GitHub requires Browser Use to be enabled in the client and the Chrome extension to report Connected. Record client version, browser profile, allowed sites, and permission level so a temporary signed-in state is not mistaken for permanent capability.

Revocation matters too. After disabling the feature or removing site permission, the agent should lose the path. If access continues, stop expansion and inspect the configuration.

Step three: define escalation actions

Treat sign-in, upload, form submission, comments, purchase, deletion, production systems, and external messages as escalation actions. The agent must pause first and show the target, fields, impact, and rollback.

For debugging, first ask the agent to reproduce the flow and identify the first divergence, not immediately edit or submit. A passing test does not prove the same behavior on customer pages, other browsers, or enterprise policies.

A minimal desk represents fixed pages, tab scope, and revocation checks. Contextual stock photo, not a browser extension or production environment. Image: Unsplash.
A minimal desk represents fixed pages, tab scope, and revocation checks. Contextual stock photo, not a browser extension or production environment. Image: Unsplash.

Step four: record failures and rollback

Record page URL, tab, step, expected and actual behavior, screenshot, permission state, and stop point. End the run on a wrong window, unexpected redirect, or sign-in challenge; the agent must not try to bypass it.

Completion requires repeatable read-only tasks, pauses before escalation, revocable permission, and evidence plus manual fallback for failures. This article did not run the extension or validate a browser version.

FAQ

Is Browser Use the same as a normal browser test script?

Not exactly. It works inside Copilot workflows and browser context, but permissions, signed-in state, determinism, and rollback still require separate validation.

Does Connected authorize form submission?

No. Connected reports extension state; submitting a form still requires separate authorization based on task scope and impact.

Fact evidence

  • Browser use entered public preview in Copilot CLI, the Copilot app, VS Code, and JetBrains when paired with the GitHub Copilot Chrome extension. — GitHub Changelog — Let GitHub Copilot interact with your browser
  • Users must enable browser use in the client and confirm that the extension reports Connected. — GitHub Changelog — Let GitHub Copilot interact with your browser
  • The feature can control a browser, debug web applications, automate browser tests, and handle web tasks. — GitHub Changelog — Let GitHub Copilot interact with your browser
  • The extension settings provide permission levels and experimental feature controls. — GitHub Changelog — Let GitHub Copilot interact with your browser

Related tools and reading

These production pages support the same decision, workflow, or safety boundary.

Conclusion

The first meaningful Browser Use receipt is not “it can click a page,” but that read-only paths, escalation, and revocation are verifiable. Make the permission boundary work before expanding automation.

What this means for everyday users

Record page URL, tab, step, expected and actual behavior, screenshot, permission state, and stop point. End the run on a wrong window, unexpected redirect, or sign-in challenge; the agent must not try to bypass it.

Related reading

Copilot App Shows AI Coding Moving from Plugins to Desktop Agents

From a global AI news perspective, GitHub Copilot App becoming available to every Copilot plan is a signal about how AI coding interfaces are evolving. The competition is no longer only about editor completions, chatbots, or benchmark headlines. It is moving toward desktop sessions, parallel task execution, BYOK model choices, GitHub workflow integration, and recurring automations. For Chinese users, the important question is not just which model is popular. It is which product can make repository permissions, account plans, model sources, task boundaries, review, and rollback clear enough for real work, especially when small teams want faster output without losing control of code and data.

OpenAI launches GPT-6 Astra with stronger computer use and explicit enterprise enablement

OpenAI introduced GPT-6 Astra on September 3 with major upgrades in computer use, browsing, software engineering, science, and professional work. The model is rolling out in phases to ChatGPT plans and is also available through the OpenAI API, Microsoft Azure, and AWS Bedrock. Enterprise access is off by default at launch and must be enabled by an administrator. OpenAI lists standard API pricing of $10 per million input tokens and $50 per million output tokens, with separate cache rates. It also classifies Astra at the Critical cybersecurity capability threshold and applies stronger safeguards. Teams should treat the reported benchmarks as vendor evidence, then run their own task, permission, latency, cost, and rollback tests before broad deployment.

GitHub Copilot Assisted Permissions Is GA: Which Requests Still Need Human Approval?

GitHub made assisted permissions generally available on October 26, 2026 across Copilot CLI, the GitHub Copilot app, and Visual Studio Code. When Copilot asks to run a command, access a file-system path, or open a URL, a language model evaluates the request. Requests classified as clearly safe can proceed automatically, while other requests still require user approval. Administrators can choose a default permission mode for new sessions, and a session falls back to manual approval when the default is unavailable or prohibited by policy. This article gives users and administrators a practical acceptance test with one harmless request, one sensitive request, policy fallback, logs, and a stop condition.

From Chat Boxes to Personal AI Companions: AI Assistants Are Entering the Desktop Execution Era

AI assistants are moving from answering questions toward continuing real tasks. AI agents, MCP tool ecosystems, personal memory, and local workbenches are pushing this shift together. For users, the real value is not another chat box, but less repeated context setup and more continuity from thinking to doing.

Gemini 4 Neon Reaches GitHub Copilot: Compare Cost per Accepted Result, Not List Price Alone

GitHub announced Gemini 4 Neon as generally available in Copilot on November 18, 2026 for Pro+, Max, Business, and Enterprise users. It uses provider list pricing under usage-based billing, administrators can manage access through model policy, and availability is rolling out gradually. This article proposes a same-task comparison instead of a list-price decision. Freeze three representative tasks and their inputs, then record accepted results, human edits, tool calls, elapsed time, and billed cost for Neon and the current model. If the model is missing, check plan, client, rollout, and policy before assuming an outage. Change the default route only when both outcome quality and total task cost meet the baseline. No model run, benchmark, or bill was performed for this article.

GitHub Copilot Evals Enters Preview: Regression-Test Skills, Agents, and MCP Changes

GitHub launched Copilot Evals in public preview on November 17, 2026. It creates repeatable tests for skills, custom agents, and MCP tools and lets users compare runs before and after a customization change. This guide defines a minimum regression set: choose three realistic tasks, then specify the input, allowed tools, success criteria, and prohibited behavior for each. Preserve at least one baseline run and one post-change run, including the result, execution path, failed checks, and human review. Merge a customization only when the change preserves the original task, permission boundary, and evidence chain. A single passing answer is not enough because model output can vary, and an automated check cannot decide every domain-specific quality or safety question.