Nature, Published online: 25 February 2026; doi:10.1038/d41586-026-00374-6
网传的漏洞演示视频，需要用户主动要求 AI 查看恶意邮件或恶意短信，才会触发攻击。如果没有用户指令，AI 并不会去自动执行高风险操作。针对视频演示的攻击方法，豆包手机助手已升级了相应的防护措施。
，详情可参考爱思助手下载最新版本
无限并行扇出 —— 一次指令，多个 Agent（Claude, Gemini, Codex, Qwen 等）同时响应（并行）
At least 57 of those that will miss the deadline are aiming to launch their service for all households by the end of 2026. More than a dozen could not give an approximate start date.
Anthropic did not immediately respond to Engadget's comment request. Earlier in the day, a spokesperson for the company said the contract Anthropic received after CEO Dario Amodei outlined Anthropic's position made “virtually no progress” on preventing the outlined misuses.