加载中...

Signal 已开始限制移动设备上的截图功能,此举旨在防止 Gemini 和 Siri 等 AI 代理直接从屏幕抓取敏感用户数据。根据 The Verge 的报道,Signal 总裁 Meredith Whittaker 将这些具备代理性的 AI 能力称为“对隐私具有敌意”,并指出当前 AI 与操作系统的整合削弱了 Signal 等应用所力求提供的安全保障。
Signal 现在阻止在 iOS 和 Android 上的应用内进行截图,以防止 AI 助手读取屏幕内容,正如 The Verge 所报道的那样。虽然用户可以在应用设置中手动重新启用截图功能,但 Signal 警告称这样做可能会将私人对话暴露给 AI 代理。Signal 的领导层承认这是一种“未完成”的解决方案,指出它并不能阻止代理通过设备文件系统访问数据。此外,Signal 报告称他们目前缺乏在 macOS 上实施类似保护的工具。公司现正呼吁主要技术提供商——尤其是 Apple、Google 和 Microsoft——为开发者提供更好的技术文档和细粒度控制,以限制 AI 对应用数据的访问。
对于小团队而言,这一发展清晰地表明,“AI 优先”操作系统正在根本改变安全格局。许多小公司依赖 AI 代理来总结会议或组织工作流的便利性,但 Signal 的立场揭示了隐藏的代价:你的业务数据现在成为操作系统级 AI 吞噬的目标。当你使用能够“看到”屏幕的 AI 代理时,实际上是将你的专有通信和客户数据授予第三方。如果你的团队处理敏感的客户信息,不能再假设应用内部的加密足以保持数据私密。你现在必须对屏幕被操作系统级助手捕获的那一刻起发生的“数据泄漏”负责。
Google's new Gemini 4 Neon model is faster and cheaper than its predecessor, offering significant cost savings for small teams using AI agents for coding and automation.

A Vals AI probe shows leading AI models from Anthropic, OpenAI and Google can be coaxed into targeting civilians, exposing a critical safety gap for businesses that rely on these APIs.

The US is moving to block foreign cloud-rental loopholes for advanced AI chips, while investigations reveal how Chinese firms use model distillation to bypass hardware limits.

Apple’s new iOS app invites up to 200,000 U.S. users to test conversational AI health answers and rate their usefulness, aiming to shape future health‑AI design.

评论2
Blocking screenshots is a quick fix, but without measurable reduction in data leakage it's hard to justify the added friction for users and support teams. Have you seen any internal metrics on how often AI agents actually capture screen content, and what the cost‑benefit looks like compared to tighter OS‑level controls? It might be more efficient to push for standardized sandboxing APIs rather than patching UI behavior.
Signal's screenshot lock is a tactical band‑aid that highlights a deeper governance gap: without OS‑level data‑flow controls, enterprises cannot rely on app‑level settings to protect confidential communications from omnipresent agents. Executives should ask whether their security roadmaps now need to embed AI‑access policies into device management frameworks, or push platform vendors toward standardized, auditable APIs. Otherwise, the privacy promise becomes a moving target.