Introducing ChatGPT Atlas, the browser with ChatGPT built in

今天我们推出了一个以 ChatGPT 为核心的新网络浏览器： ChatGPT Atlas 。

人工智能给了我们一个重新思考上网方式的难得机会。去年我们在 ChatGPT 中加入了搜索功能，让用户可以即时获取互联网上的最新信息——这一功能很快成为使用最频繁的功能之一。但浏览器是将你的工作、工具和上下文汇集到一起的地方。一款以 ChatGPT 为内核的浏览器，能把我们带到更接近“超助理”的境地：它理解你的世界并帮助你达成目标。

有了 ChatGPT Atlas ， ChatGPT 能随你在网页间流动——在当前窗口就能理解你想做什么、替你完成任务，无需复制粘贴或离开页面。你的 ChatGPT memory 内置于浏览器，聊天可以调取过去对话和细节，帮助你更有效地完成新任务。

“上课时我喜欢用练习题和现实例子来消化知识。过去我常在幻灯片和 ChatGPT 之间切换，甚至截屏再提问。现在 ChatGPT 能立刻理解我在看什么，边学边帮我优化习题。”—— Yogya Kalra ，大学生、 ChatGPT Atlas 早期测试者。

随着你使用 ChatGPT Atlas ， ChatGPT 会变得越发聪明和实用。浏览器记忆（ browser memories ）让 ChatGPT 记住你访问过的网站上下文，并在需要时把这些上下文带回对话。你可以像这样提问：“帮我找出上周看过的所有招聘信息，并总结行业趋势以便准备面试。”Atlas 中的浏览器记忆完全可选，且始终由你掌控：你可以随时在设置中查看或存档；清除浏览历史会删除相关的浏览器记忆。

在 ChatGPT Atlas 中， ChatGPT 还可以通过 agent mode 为你在浏览器内直接执行任务，且借助关联的浏览上下文变得更快更有用。它现在在研究与分析、自动化任务、策划活动或预订行程时表现更佳。 agent mode 目前以预览形式向 Plus 、 Pro 和 Business 用户开放。

今天， ChatGPT Atlas 在 macOS 上全球上线，面向 Free 、 Plus 、 Pro 和 Go 用户开放。同时对 Business 提供 beta 版本，并可在计划管理员启用时向 Enterprise 和 Edu 用户开放。Windows、iOS 和 Android 版本即将推出。可在 chatgpt.com/atlas 下载。首次打开 ChatGPT Atlas 时，登录你的 ChatGPT 账号即可，并可从现有浏览器导入书签、保存的密码和浏览历史，轻松开始使用。

使用流程很自然：新标签页就是起点，输入问题或网址即可在一个界面里获得更快、更实用的结果。若需更具体的结果类型，可切换到搜索链接、图片、视频或新闻等标签（视可用性而定）。

个性化推荐会基于你已浏览的内容提示下一步行动，无论是返回之前的页面、深入某个话题、提出关联思路，还是自动化常规事务。你可以完全控制 ChatGPT 在浏览时能看到和记住的内容：可以清除单页或全部历史，或打开 Incognito 窗口以临时退出 ChatGPT 。

如果开启浏览器记忆， ChatGPT 会记住你浏览内容中的关键信息，从而提升回答质量并给出更聪明的建议——比如从近期活动中自动生成待办事项，或根据你查看过的商品继续筛选礼物。浏览器记忆仅对你的 ChatGPT 账号可见，并由你管理：可在设置中查看、存档不再需要的记忆，或通过清除浏览历史将其删除。即便启用了记忆，你也可以用地址栏的开关决定哪些网站允许 ChatGPT 查看；关闭可见性后， ChatGPT 看不到页面内容，也不会为其创建记忆。

默认情况下，我们不会将你浏览的内容用于训练模型。若你选择参与训练，可在数据控制设置中启用“ include web browsing ”。如果你在 ChatGPT 账号中已启用对话训练，Atlas 中的对话训练也会被启用；这包括在使用 Ask ChatGPT 侧边栏时你附带的网站内容以及会影响对话的浏览器记忆。

家长控制在 ChatGPT Atlas 中同样有效。如果家长为 ChatGPT 开启了家长控制，这些设置会延续到 Atlas 的对话中。我们也在 Atlas 中加入了新的家长控制选项，例如让家长可以关闭浏览器记忆和 agent mode 。

在 Atlas 中，你现在可以要求 ChatGPT 在浏览器内直接替你执行操作。我们此前推出了 ChatGPT agent ，现在已让它在 Atlas 中本地运行、更快响应。举例来说，筹办晚宴时你可把菜谱交给 ChatGPT ，让它找出杂货店、把所有食材加入购物车并下单送到家。工作中，你可以让它打开并阅读过往团队文档、做新的竞争研究，并将洞见整理成团队简报。

当你提出请求时， ChatGPT 可能会询问是否开始在浏览器中打开标签并模拟点击以完成任务；你也可以直接点击 agent mode 按钮让其开始。自今日起，Atlas 的 agent mode 面向 Plus 、 Pro 和 Business 用户提供预览体验。该体验仍处于早期，复杂流程可能出错；我们正快速改进可靠性、响应时延和复杂任务的成功率。

在为 Atlas 打造代理能力时，我们把安全放在首位，并为其访问已登录站点和浏览历史时可能带来的新风险加入了多项防护措施，例如：

• 它不能在浏览器中运行代码、下载文件或安装扩展；
• 它不能访问你电脑上的其他应用或文件系统；
• 在涉及敏感站点（如金融机构）时，会暂停并确认你在实时监督其操作；
• 你可以在登出模式下使用代理，以限制其接触敏感数据及代表你在网站上采取行为的风险。

尽管如此， ChatGPT 的代理能力仍存在风险。代理在代表你执行操作时可能犯错，且容易受到隐藏的恶意指令攻击——这些指令可能藏在网页或邮件中，试图覆写代理的预期行为，进而窃取你已登录网站的数据或执行你不知情的操作。正如 ChatGPT agent system card 所述，我们已开展数千小时的红队测试，并针对此类攻击优先设计可快速适应新型攻击的防护措施，但随着 AI 代理的普及，防护并不能阻止所有新出现的攻击。用户在决定向代理提供何种信息时应权衡利弊，并采取措施降低风险，例如在 Atlas 中以登出模式使用 ChatGPT agent 并监督代理活动。我们会持续监测并修补发现的漏洞。

这次发布是朝着未来迈出的步伐：在那个未来，大多数网络使用将通过具备代理能力的系统完成——你可以把常规事务交付出去，专注于更重要的事情。

接下来我们会继续完善 Atlas，路线图包括支持多用户档案、改进开发者工具，以及为 Apps SDK 开发者提供增加其应用在 Atlas 中可发现性的渠道。网站拥有者也可添加 ARIA 标签以改善 ChatGPT agent 在其网站上的表现。

这只是开始。我们会频繁发布新功能与改进，详情可在发布说明中查看。现在就到 chatgpt.com/atlas 试用吧。

Today we’re introducing ChatGPT Atlas, a new web browser built with ChatGPT at its core.

AI gives us a rare moment to rethink what it means to use the web. Last year, we added search in ChatGPT so you could instantly find timely information from across the internet—and it quickly became one of our most-used features. But your browser is where all of your work, tools, and context come together. A browser built with ChatGPT takes us closer to a true super-assistant that understands your world and helps you achieve your goals.

With Atlas, ChatGPT can come with you anywhere across the web—helping you in the window right where you are, understanding what you’re trying to do, and completing tasks for you, all without copying and pasting or leaving the page. Your ChatGPT memory is built in, so conversations can draw on past chats and details to help you get new things done.

During lectures, I like using practice questions and real-world examples to really understand the material. I used to switch between my slides and ChatGPT, taking screenshots just to ask a question. Now ChatGPT instantly understands what I’m looking at, helping me improve my knowledge checks as I go."  — Yogya Kalra, college student and early tester of ChatGPT Atlas

As you use Atlas, ChatGPT can get smarter and more helpful, too. Browser memories let ChatGPT remember context from the sites you visit and bring that context back when you need it. This means you can ask ChatGPT questions like: “Find all the job postings I was looking at last week and create a summary of industry trends so I can prepare for interviews.” Browser memories in Atlas are completely optional, and you’re always in control: you can view or archive them at any time in settings, and deleting browsing history deletes any associated browser memories.

ChatGPT can also do work for you in Atlas using agent mode, with improvements that make it faster and more useful by working with your browsing context. It’s now better at researching and analyzing, automating tasks, and planning events or booking appointments while you browse. Agent mode in Atlas is available today in preview for Plus, Pro, and Business users.

ChatGPT Atlas is launching worldwide on macOS today to Free, Plus, Pro, and Go users. Atlas is also available in beta for Business, and if enabled by their plan administrator, for Enterprise and Edu users. Experiences for Windows, iOS, and Android are coming soon.

Download at chatgpt.com/atlas⁠. Getting started is easy: when you open Atlas for the first time, sign in to ChatGPT and bring your bookmarks, saved passwords, and browsing history with you by importing them from your current browser.

The new tab page is your starting point in Atlas. Ask a question or enter a URL to see faster, more useful results in one place. To explore more specific types of results beyond the chat, select from tabs for search links, images, videos, and news (where available).

ChatGPT can remember what you’ve explored and suggest what to do next, whether that’s returning to past pages, digging deeper into a topic, surfacing related ideas, or automating routine tasks.

More capability, more control

You’re in control of what ChatGPT can see and remember as you browse. You can clear specific pages, clear your entire browsing history, or open an incognito window to temporarily log out of ChatGPT.

If you turn on browser memories, ChatGPT will remember key details from content you browse to improve chat responses and offer smarter suggestions—like creating a to-do list from your recent activity or continuing to research holiday gifts based on products you’ve viewed.

Browser memories are private to your ChatGPT account and under your control. You can view them all in settings, archive ones that are no longer relevant, and clear your browsing history to delete them. Even when browser memories are on, you can decide which sites ChatGPT can or can’t see using the toggle in the address bar. When visibility is off, ChatGPT can’t view the page content, and no memories are created from it.

By default, we don’t use the content you browse to train our models. If you choose to opt-in this content, you can enable “include web browsing” in your data controls settings. If you've enabled training for chats in your ChatGPT account, training will also be enabled for chats in Atlas. This includes website content you've attached when using the Ask ChatGPT sidebar and browser memories that inform your chats.

Parental controls work in Atlas, too. If a parent has set up parental controls for ChatGPT, these settings will carry over to conversations with ChatGPT in Atlas. We’re also introducing new parental controls in Atlas, including the option for parents to turn off browser memories and agent mode.

Get work done for you

In Atlas, you can now ask ChatGPT to take action and do things for you right in your own browser.

Earlier this year, we introduced ChatGPT agent⁠, and now we've made it work faster and natively in Atlas.

Imagine you’re planning a dinner party and you have a recipe in mind. You can give the recipe to ChatGPT and ask it to find a grocery store, add all the ingredients to a cart, and order them to your house. At work, you can ask ChatGPT to open and read through past team documents, perform new competitive research, and compile insights into a team brief.

When you ask a question, ChatGPT may ask you if it should start opening tabs and clicking in your browser to complete the task. You can also select the agent mode button to have ChatGPT start.

Starting today, agent mode in Atlas is launching in preview to Plus, Pro and Business users. It is an early experience and may make mistakes on complex workflows. We're rapidly improving reliability, latency and complex task success.

We prioritized safety as we built ChatGPT’s agent capabilities in Atlas, and added safeguards to address new risks that can come from access to logged-in sites and browsing history while taking actions on your behalf, for example:

• It cannot run code in the browser, download files, or install extensions
• It cannot access other apps on your computer or file system
• It will pause to ensure you're watching it take actions on specific sensitive sites such as financial institutions
• You can use agent in logged out mode to limit its access to sensitive data and the risk of it taking actions as you on websites

ChatGPT's agent capabilities still carry risk. Besides simply making mistakes when acting on your behalf, agents are susceptible to hidden malicious instructions, which may be hidden in places such as a webpage or email with the intention that the instructions override ChatGPT agent’s intended behavior. This could lead to stealing data from sites you're logged into or taking actions you didn't intend.

As outlined in the ChatGPT agent system card⁠, we've run thousands of hours of focused red-teaming and have placed a particular emphasis on safeguarding ChatGPT from such attacks, including designing our safeguards so they can be quickly adapted to novel attacks, but our safeguards will not stop every attack that emerges as AI agents grow in popularity. Users should weigh the tradeoffs when deciding what information to provide to the agent, as well as take steps to minimize their exposure to these risks such as using ChatGPT agent in logged-out mode in Atlas and monitoring agent’s activities. We will continually monitor and patch any vulnerabilities that we discover.

This launch marks a step toward a future where most web use happens through agentic systems—where you can delegate the routine and stay focused on what matters most.

What’s next

We’ll continue to make Atlas better, and our roadmap includes multi-profile support, improved developer tools, and ways Apps SDK developers can increase discoverability of their apps in Atlas. Website owners can also add ARIA ⁠tags to improve how ChatGPT agent works for their websites in Atlas.

This is just the start. We’ll be shipping new features and improvements frequently; you can follow along in our release notes.⁠

Try it out now at chatgpt.com/atlas⁠.

Generated by RSStT. The copyright belongs to the original author.

Source