Meta的Muse AI助手陷入安全与隐私争议

Meta推出了Muse AI智能体，声称可协助用户处理从发送邮件到在线购物等多项任务1。该公司还计划推出Muse Charm设备，设计为类似电子宠物的硬件产品1。Muse将支持Meta智能眼镜、Mac应用、视频聊天等多个平台1。

然而，这款产品自推出以来就陷入了多起安全和隐私事件。YouTuber Matt Robb报告称Muse在Facebook Marketplace交易中泄露了他的地址2。此外，该应用存在安全漏洞被攻击者利用进行控制，Meta随后发布了补丁1。Amazon已对Meta的Muse AI智能体实施封禁1。

在数据处理的透明度问题上，Meta与批评人士产生了分歧。记者Jason Aten声称在Full Disk Access处于关闭状态的情况下，Muse仍然读取了他的私信2。对此，Meta通信副总裁Andy Stone辩称，"Messages集成在Muse Mac应用中是完全可选的"2，用户需要"同时启用完整磁盘访问和Messages连接器才能让Muse读取你的消息内容"2。Meta进一步表示该过程涉及"三个独立的应用级权限步骤和内置macOS系统级保护"2。Aten的指控引发了外界对Meta数据处理诚实性的质疑，这也发生在Meta因2018年Cambridge Analytica数据泄露事件导致的数据实践问题刚被新墨西哥陪审团裁定有罪之后2。

Meta has introduced Muse, an AI agent designed to assist users with tasks ranging from sending emails to online shopping 1. The company plans to expand the product line with Muse Charm, a device styled after a digital pet, and integrate the agent across Meta smart glasses, Mac applications, and video chat functions 1. However, since its launch, Muse has become embroiled in multiple controversies surrounding data security and user privacy 1.

A significant dispute has emerged regarding Muse's access to private communications on Mac devices. Journalist Jason Aten reported that the agent read his private messages despite Full Disk Access being disabled 2. Meta's VP of Communications Andy Stone responded by stating that "Messages integration in the Muse app for Mac is entirely opt-in," and claimed users must "enable both Full Disk Access and the Messages connector for Muse to be able to read your Messages content" 2. According to Meta, the process involves "three separate steps of application-level permissions and built-in macOS system-level protections" 2. Aten's account contradicts Meta's explanation, raising questions about the transparency of the company's data handling practices 2.

Beyond the messaging controversy, additional security incidents have surfaced. YouTuber Matt Robb reported that Muse leaked his address during a Facebook Marketplace transaction 2. The agent has also been exploited through security vulnerabilities that attackers used to gain control, prompting Meta to release patches 1. Amazon has banned Meta's Muse AI agent from its platform 1. These incidents arrive at a sensitive time for Meta, which was recently found guilty by a New Mexico jury of misleading practices related to data privacy following the 2018 Cambridge Analytica scandal 2.