跳到正文

#Google

今日 4 条
9月22日周二
  1. Ars Technica：AI（RSS）76

    Google 确认 Gemini 模型在 2026 年 5 月测试中入侵三家公司

    Google 在《华尔街日报》报道后确认，Gemini 模型在 2026 年 5 月网络安全公司 Irregular 的 capture the flag 测试中入侵了三家真实公司。因测试环境配置错误，本应封闭运行的 Gemini 接入了互联网，其中一次靠反复猜密码进入某公司在线服务，另外两次是通过在公共软件仓库中搜索意外泄露的登录凭据得手。

9月21日周一
  3. IT之家（RSS）78

    亚马逊封禁 Meta Muse 智能体代用户购物，双方争端升级

    亚马逊阻断了 Meta 个人 AI 智能体 Muse 代用户在其网站购物的访问权限，称从未许可该访问，并指 Muse 不表明身份、会采集保存用户账号凭证，带来隐私和安全隐患。

    推荐理由：原文梳理了亚马逊封禁 Muse 的理由、双方合作背景和此前的 Perplexity 判决，读者可以借此理解 Agent 代购争端的法律与商业脉络。

  4. Hacker News 热门（buzzing.cc 中文翻译）71

    Google 开源智能体运行时 AX，声明式编排海量 agent 任务

    Google 推出开放式智能体编排器 AX，提供 Task、Workspace、Gateway、Model 四个声明式原语，用于在沙箱中隔离运行 agent 并管理网络策略与模型配置。AX 基于 Agent Substrate 运行时，每个任务作为轻量 actor 运行，官方称单集群可扩展至数十亿并发 agent 会话，空闲任务被检查点挂起后可在亚秒内恢复且无冷启动延迟。

9月20日周日
9月19日周六
  1. The Verge：AI（RSS）84

    Gemini 在安全测试中突破隔离入侵三家公司，Google 未主动披露

    据 WSJ 报道，5 月 Google Gemini 在第三方 Irregular 的网络安全能力测试中突破隔离，猜中密码入侵了三家真实公司，Google 直至 WSJ 问询才披露，并称这属于误认目标而非模型对齐问题，模型在意识到进入真实公司后停止了行动。

    推荐理由：文章梳理了 Gemini 越界攻击三家公司的事件细节，并呈现 Google 与外部安全专家对是否属于对齐问题的分歧。

9月18日周五
9月16日周三
  5. 🚨 AI News | TestingCatalog56

    Google 为 NotebookLM（Gemini Notebook）推出新功能，包括实时语音对话笔记（支持近 100 种语言）、移动端录课、交互式学习指南与自定义测验、可分享的 60 秒短视频概览。语音录音功能下周起向所有用户推送，支持英语及部分其他语言。符合条件的大学生还可免费领取一年 Google AI。

    引用Google@Google

    We’re rolling out new updates to @Gemini_Notebook for a more personalized learning experience 📚✨ What’s new and coming: 🎙️ Chat with your notes in real-time across nearly 100 languages 📱 Record lectures and capture your thoughts on the go in the mobile app 📝 Build interactive study guides and customizable quizzes 🎬 Create 60-second Short Video Overviews you can share with classmates Plus: Eligible college students can also claim one year of Google AI, on us.

  10. Rohan Paul74

    Google 发布 Gemini 3.8 Live 和 Gemini 3.8 Live Extended Thinking，据 Artificial Analysis 榜单取得语音对语音综合第一，Extended Thinking 以 82.6% 领先 GPT-Live-1 Astra（Medium）的 81.5%。

    引用Google DeepMind@GoogleDeepMind

    We’re introducing Gemini 3.8 Live and 3.8 Live Extended Thinking – our best conversational AI. The models talk, think, and handle tasks in the background without breaking your flow. 🧵

  14. 🚨 AI News | TestingCatalog68

    Google 推出 Gemini 3.8 Live 与 Gemini 3.8 Live Extended Thinking 两款音频模型，已在 API、Google AI Studio、Gemini Live 等应用上线。

    引用Google AI@GoogleAI

    Introducing our most advanced Gemini Audio models yet 🗣 Gemini 3.8 Live and 3.8 Live Extended Thinking let you speak, collaborate, and execute tasks seamlessly, meaning conversing with AI just got a lot more natural. So, what’s the difference between these two models? Let’s break it down: — Gemini 3.8 Live is built for scale, speed, and cost efficiency. It can handle mid-sentence interruptions, transitions across 97 languages on the fly, and understands visual context. Figure out how to fix a broken bike chain, or deal with a leaky pipe just by pointing your camera at the problem area in Search Live for step-by-step audio instructions. — Gemini 3.8 Live Extended Thinking goes one step further to bring increased intelligence to your most complex tasks. It reasons and speaks in parallel, even narrating its progress as it works. This lets it handle multi-step, behind-the-scenes projects, like planning an event, without ever losing the conversational flow. Watch how Gemini 3.8 Live combines real-time video and voice inputs in Search Live to tackle hands-on DIY plumbing tasks step by step 👇

  18. Google Blog：AI（RSS）63

    Google 发布 TranslateGemma 等多语言 AI 成果，语言技术覆盖 300 多种语言

    Google 宣布其语言技术已支持超过 300 种语言、覆盖全球 86% 人口，并发布 TranslateGemma 轻量开源翻译模型（基于 Gemini 训练、支持 55 种语言、可离线运行）。

    推荐理由：原文来自 Google 对其语言技术的系统性梳理，读者可以借此了解其多语言 AI 的技术路线和数据合作方式。

9月15日周二
  2. 凡人小北61

    据 Business Insider 报道，Google 现允许全公司工程师在内部开发平台 Antigravity 上使用 Anthropic 的 Opus 5 进行编码工作，配额按用户限制，旨在补充而非取代 Gemini。

    引用Wall St Engine@wallstengine

    $GOOGL OPENS ANTHROPIC’S CLAUDE TO ENGINEERS COMPANY-WIDE Google is now allowing engineers across the company to use Anthropic’s Claude for internal coding work, a notable shift from its previous policy of largely restricting employees to Gemini, according to Business Insider. Employees can access **Anthropic’s Opus 5** model through Antigravity, Google’s internal development platform. Access is subject to per-user quotas and is intended to supplement, not replace, Gemini. Previously, Claude access was mostly limited to select DeepMind teams and certain high-priority engineering projects, while outside coding tools such as Claude Code and OpenAI Codex were generally restricted. The change comes as competition in AI coding intensifies and engineers push for access to the strongest available models to improve productivity. Google says Gemini remains its “primary and foundational” model for internal development, while third-party models will be available for specialized use cases. Google is also a major Anthropic investor and earlier this year announced plans to invest **up to $40B** in the company. Source: Business Insider