Meta launches AI agent that can access other apps to send emails, make payments

Meta’s Muse agent is designed to access a person’s apps across categories like email, calendar, payments, health, shopping and the smart home.

Published Sep 09, 2026, 03:52 AM

Updated Sep 09, 2026, 03:54 AM

Meta launched Muse, an AI agent that can autonomously send emails, make payments, and book travel, initially available in the US via a dedicated app and WhatsApp.

Muse accesses users' apps across categories like email, calendar, payments, and smart home, running on virtual machines for continuous background operation.

Internal tests showed mixed results, with useful features but also security flaws and reliability issues, prompting Meta to delay release for improved safety and privacy measures.

NEW YORK - Meta rolled out on Sept 8 a long-touted AI assistant that can autonomously send emails, sell a car and book travel on a person’s behalf, despite internal concerns that the technology mismanages its access to sensitive personal data.

The company’s Muse agent, known internally as Hatch, is the centrepiece of chief executive officer Mark Zuckerberg’s plan to offer “personal superintelligence” to the billions of people who use Meta’s services daily.

The product will be available only in the US initially, via a dedicated Muse app or Meta’s WhatsApp messaging service, Meta said in its announcement.

Meta said it plans to add the agent to its line of smart glasses “soon,” without elaborating.

Modelled on the open-source AI agent OpenClaw, Muse is designed to access a person’s apps across categories like email, calendar, payments, health, shopping and the smart home, Meta said. People choose which apps it connects to and can revoke access at any time.

Each Muse agent runs on its own virtual machine, a cloud-based emulation of a personal computer, which enables it to keep carrying out requests in the background even when a person is not actively using it.

Syncing up with apps containing a person’s real data increases the agent’s potential utility, while also significantly raising the stakes for safety and reliability issues, both for the users who have entrusted it with their information and others who may be on the receiving end of agent misbehavior.

Vishal Shah, vice-president of AI products at Meta, said the company had initially delayed the release of the product in April to make it more secure. Meta determined that work had allowed it to “cross the threshold” and meet its minimum requirements for product safety, security, privacy, model performance and other metrics.

“It is impossible to say that there is never going to be a mistake, but every single part of the architecture has been designed to make this as safe, as secure, as private as we can possibly make it,” Shah said.

Internal tests reveal mixed results

As recently as this week, Meta employees testing the tool internally have reported cases of Muse successfully arranging vacation logistics, disconnecting without explanation and uploading sensitive information without permission, according to internal posts seen by Reuters.

One person wrote that the product had been so useful in arranging itineraries and ground transportation for them that they described it as “the third participant” on their recent three-week honeymoon in Indonesia.

In another post, an employee who had prompted Muse to monitor for tickets and other items that sell out quickly reported encountering “many failure modes that made it unreliable.” The product stopped refreshing the page after about 15 minutes, silently ignored other errors and at times disabled monitoring “for no apparent reason,” the person said.

Meta chief technology officer Andrew Bosworth posted that he kept getting logged out and needing to log back in, sometimes several times within a few minutes.

Others flagged serious security flaws, like an agent routing around guardrails to expose a person’s personal iCloud photos after being prompted to identify toys visible in pictures from a child’s birthday party.

Meta did not immediately respond to a request for comment on the specific incidents. REUTERS

AI/artificial intelligence

Meta 的 Muse 代理旨在访问用户在电子邮件、日历、支付、健康、购物和智能家居等各个类别中的应用程序。

发布于 2026 年 9 月 9 日 上午 3:52

更新于2026年9月9日凌晨3:54

Meta推出了Muse，这是一款人工智能代理，可以自主发送电子邮件、进行支付和预订旅行，最初通过专用应用程序和WhatsApp在美国提供服务。

Muse 可访问用户的电子邮件、日历、支付和智能家居等各类应用程序，并在虚拟机上运行以实现持续的后台运行。

内部测试结果喜忧参半，虽然有一些有用的功能，但也存在安全漏洞和可靠性问题，这促使 Meta 推迟发布，以改进安全性和隐私措施。

纽约——尽管内部有人担心该技术在访问敏感个人数据方面管理不善，但 Meta 公司还是在 9 月 8 日推出了一款备受瞩目的 AI 助手，该助手可以自主发送电子邮件、出售汽车和代表用户预订旅行。

该公司名为 Muse 的智能代理（内部称为 Hatch）是首席执行官马克·扎克伯格计划的核心，该计划旨在为每天使用 Meta 服务的数十亿人提供“个人超级智能”。

Meta 在公告中表示，该产品最初将仅在美国发售，可通过专门的 Muse 应用或 Meta 的 WhatsApp 即时通讯服务购买。

Meta表示计划“很快”将该代理添加到其智能眼镜产品线中，但没有详细说明。

Meta表示，Muse基于开源人工智能代理OpenClaw建模，​​旨在访问用户的各种应用程序，涵盖电子邮件、日历、支付、健康、购物和智能家居等领域。用户可以选择允许Muse连接哪些应用程序，并可随时撤销访问权限。

每个 Muse 代理都运行在自己的虚拟机上，虚拟机是基于云的个人电脑模拟器，即使用户不主动使用它，它也能继续在后台执行请求。

与包含个人真实数据的应用程序同步，可以提高代理的潜在效用，但同时也大大增加了安全性和可靠性问题的风险，这不仅对那些将信息委托给代理的用户而言如此，对那些可能因代理不当行为而受到影响的人来说也是如此。

Meta公司人工智能产品副总裁Vishal Shah表示，公司最初推迟了该产品在4月份的发布，目的是为了提高其安全性。Meta公司认为，经过改进，该产品已经“跨越了门槛”，达到了其在产品安全、隐私、模型性能和其他指标方面的最低要求。

“我们无法保证绝对不会出错，但建筑的每一个部分都经过精心设计，力求使其尽可能安全、可靠、私密，”沙阿说。

内部测试结果喜忧参半

据路透社看到的内部帖子显示，就在本周，Meta 内部测试该工具的员工报告称，Muse 成功安排了度假行程，在没有解释的情况下断开连接，并在未经许可的情况下上传敏感信息。

一位用户写道，该产品在为他们安排行程和地面交通方面非常实用，以至于他们将其描述为他们最近在印度尼西亚为期三周的蜜月旅行中的“第三位参与者”。

在另一篇帖子中，一位曾要求 Muse 监控门票和其他快速售罄商品的员工报告称，他遇到了“许多故障模式，导致该功能不可靠”。该员工表示，该产品会在大约 15 分钟后停止刷新页面，默默忽略其他错误，有时还会“无缘无故”禁用监控功能。

Meta 首席技术官 Andrew Bosworth 发帖称，他经常被登出，需要重新登录，有时几分钟内就要登出好几次。

其他人则指出了严重的安全漏洞，例如，在被要求识别儿童生日派对照片中出现的玩具后，代理绕过防护措施，泄露了用户的个人 iCloud 照片。

Meta公司尚未就具体事件作出回应。路透社

人工智能