Signal 非常重视保护你的隐私
Restricting AI agents in mobile operating systems platforms from capturing data

原始链接: https://signal.org/blog/protecting-your-privacy/

Signal 正在通过在 iOS 27 上屏蔽截图,以及在 AI 助手尝试在 Android 上进行屏幕捕获时进行拦截,来限制操作系统级 AI 代理获取其数据;此前,它针对 Windows 上的 Microsoft Recall 采取了类似措施。Google、Apple 和 Microsoft 正在将代理式 AI 深度嵌入其操作系统,这些代理会扫描屏幕内容——包括 Signal 消息——通过在 Signal 控制范围之外捕获、存储和暴露敏感数据,削弱 Signal 的隐私保障。用户可以在设置中关闭该保护,但这会危及用户及其联系人的数据。Signal 指出,这一补救措施并不完整,因为 AI 代理仍可能通过文件系统等其他方式访问数据,而开发者也缺乏强大工具或文档来阻止此类访问。Signal 呼吁 Apple、Google 和 Microsoft 提供易于使用且文档完善的 AI 数据访问退出机制,并改进技术文档,警告称若无这些保护,操作系统级 AI 集成将对应用层隐私构成生存威胁。Signal 还正与操作系统提供商及盟友合作,推动行业范围的隐私标准。

Hacker News 的帖子讨论了 Signal 阻止内置移动操作系统 AI 代理捕获和存储 Signal 数据的努力。用户指出,操作系统层面的 AI 可以读取屏幕内容,即使聊天已加密,也会带来隐私风险。提到了 Apple 的 Siri 截图访问权限以及 EFF 推动应用级排除设置。评论者争论 Signal 是否应做更多:支持 GrapheneOS 等隐私操作系统,允许用户控制 AI 访问,或向用户警示间谍软件。其他人质疑 Signal 更广泛的隐私声明,理由包括电话号码/账户要求、元数据收集、云备份和集中式基础设施。一些人认为操作系统层面的控制是必要的;另一些人则说应由用户而非 Signal 决定。总体而言,该帖子突显了操作系统集成 AI 与端到端隐私之间的张力,以及对 Signal 的保护和实践是否足够的分歧。
相关文章

原文

Black silhouette in the shape of a smartphone overlaid with graffiti-style text that says "SIGNAL WAS HERE".

Privacy is our mission; it’s why we exist, and we do everything we can to protect it.

This means that we are restricting the latest AI agents built into mobile operating systems from potentially capturing and storing your Signal data.

We’ve already had to do this once with Windows. Last year, we used screenshot blocking to prevent Microsoft’s Recall AI from taking and storing screenshots of your Signal messages on Windows 11 operating systems by default. We will now be taking similar measures to restrict agentic systems built into iOS 27 and on Android versions where we can detect an AI assistant attempting to read your Signal screen.

The problem is similar to Microsoft Recall on Windows: Google and Apple are integrating agentic capabilities like Gemini Intelligence with Google Gemini and Onscreen Awareness with Siri AI deep into their operating systems. This is a significant move. Google (Android), Apple (iOS/macOS), and Microsoft (Windows) produce the three major consumer operating systems (OS), which comprise the environment that third-party developers – like Signal – build their apps to function and operate within. To put it differently, these three operating systems are the oceans in which Signal, and other applications, must swim; they dictate the conditions of life for the application ecosystem. Right now, these agentic AI integrations are changing this environment in a way that’s making it increasingly hostile to privacy. This is an existential threat to Signal, and our ability to responsibly uphold our mission.

In order to gain access to information about you, these OS AI agents “scan” the information that’s on your screen. In doing so, your Signal data is captured, then made available to these agentic systems in order to allow them to answer your questions and do your bidding.

The net result is that these AI capabilities in the operating system are undermining Signal’s ability to independently guarantee privacy. For over a decade, this guarantee has been backed by our tried and trusted open source code base and our formally verified encryption protocol. These have been scrutinized by some of the best security experts and hackers across the world, and shown to be robust. Now, in service of AI agentic capabilities, your sensitive data is being captured, stored, and made accessible to whatever local database or remote server an OS-level AI agent is using at that moment. All of this is outside of Signal’s control once the data has been captured, and imperils the privacy of all the participants in a conversation.

A screenshot from Signal Settings on iOS showing the new "Block Siri AI" setting enabled.

A screenshot from Signal Settings on Android showing the "Screen security" setting enabled.

So, we are blocking such capture by preventing screenshots in the app on iOS 27 and when AI assistants attempt to capture screen data on Android.

People who want the ability to take screenshots can disable this protection within the Signal app’s settings. However, doing so will potentially expose your (and your contacts’) Signal content to these AI agents, and break our privacy guarantees. Our remedy is also incomplete – while we are doing everything we can, restricting screenshots does not address other more sneaky ways AI agents could potentially access application data, such as retrieving content already present on a user’s device via the filesystem.

We want to be very clear: it is a serious problem that Signal – and application developers beyond – lack robust tools to prevent AI agents from undermining application-level privacy. Even where we are able to detect and block AI data access, Google, Apple, Microsoft, and others do not provide ways for developers to easily restrict such AI data access without making drastic compromises to the core user experience. This is also the reason we are currently unable to implement these protections on macOS.

We are also struggling with a significant lack of documentation about how these AI agents actually work. Over the past few months, we’ve spent considerable developer time and energy doing forensic investigations across Android and iOS in an attempt to learn what is actually happening when these AI systems access application data. Without better documentation and clear technical specifications, it will be increasingly difficult for Signal to provide robust privacy and security to the many millions of people who rely on us.

We continue to call on the major consumer operating system companies – Apple, Google, and Microsoft – to provide simple technical tools and accurate documentation that give developers like Signal ways to prevent AI agents from accessing sensitive application data. In particular, we are calling on these companies to provide two common-sense remedies that can protect the privacy so many millions of people rely on:

  1. Developers should have an easy and well-documented way to opt their app out of AI data access and processing, including (but not limited to) screen reading, filesystem access, notifications, APIs, and other OS frameworks, all without undermining the core utility of an application.
  2. Operating system companies should create better documentation and technical specifications around how these AI agents function, particularly regarding data access and processing. Confidential computing approaches are not a complete solution, and do not address these core concerns. In particular, these approaches still leave the power to determine the extent of privacy guarantees (if any) in the hands of these providers.

We also see a growing need for tools that can prevent third-party agents, like Meta’s Muse, from accessing application data using the deep permissions that users are encouraged to grant them during setup. In many cases these agents are leveraging accessibility settings that were meant to empower actual people to get the most out of their apps and services, not to act as a skeleton key that provides access to AI agents.

In addition to taking these technical measures to block capture, we are engaging in conversations across the ecosystem to make the stakes clear. There is a lot of work to be done in the industry to improve the ways that AI agents operate in the operating system, and how and whether they capture and process private data. We are cautiously optimistic that some major OS providers, like Google and Apple, are engaging with us on how these systems should better interface with privacy-sensitive apps like Signal. We hope to have more to share on this topic soon, and that these companies recognize the danger to the ecosystem and work with us and others to create standardized approaches that meaningfully protect privacy across the industry.

Finally, we want to acknowledge that we are not the only ones recognizing these perils, and we are grateful to our allies for pushing for similar changes. We all need to recognize that the current integration of AI agents into the operating system, without these common-sense remedies, represents an existential threat to privacy at the application layer, and the ability of independent developers like Signal to protect your sensitive data. Signal is the canary in the coal mine, but this affects the entire ecosystem and should galvanize anyone who cares about these fundamental rights.


Signal Technology Foundation is the nonprofit home of the Signal app, which is trusted by millions around the world. We protect free expression and enable secure global communication through open source privacy technology. Signal Technology Foundation is a US 501(c)(3) tax-exempt organization.

联系我们 contact @ memedata.com