AI AI 每日简报 Daily Digest
2026年09月24日周四
最新 桑德斯提案禁止超级智能开发，违者最高判20年 The Verge
AI DAILY DIGEST 2026年09月24日 · 周四

AI 每日简报Daily Brief

覆盖窗口（Asia/Shanghai）：09-22 20:31 — 09-24 08:31　|　最近 36 小时　|　已排除往期重复 15 条
13
入选条目
49
清洗后候选
0
聚合站来源
13
源站直达补充

今日综述 / TODAY'S BRIEF

今日主线是监管压力与产品落地同时加速。桑德斯提出禁止超级智能开发的立法草案、加州强制数据中心披露水电消耗，加上盖洛普调查显示即便每日使用 AI 的美国人仍有 68% 感到担忧，说明公众与政策层对 AI 的不信任并未随普及而消退。与此同时，Meta 的 Muse、ChatGPT 移动端语音智能体和 Spotify 的 Taste Profile 把代理与个性化能力推向大众日常场景，而 Harvey、invideo、Airbnb 的案例则显示 GPT-6 Astra 正深入法律、视频后期与企业工程等专业工作流。资本层面，Ema 以 7700 万美元 B 轮融资切入企业软件自动化，其"AI 员工"模式被视为对传统 SaaS 与 IT 服务支出的直接冲击。值得注意的信号是，OpenAI 将 Daybreak 计划扩展至乌克兰政府用于民用网络防御，AI 能力正被直接引入国家级安全场景。整体看，能力扩张与监管、信任之间的张力正在成为行业最核心的变量。

The dominant thread today is the simultaneous acceleration of regulatory pressure and product deployment. Bernie Sanders' bill to ban superintelligence development and California's new data center water-and-power disclosure rules, combined with Gallup polling showing that 68% of Americans who use AI daily still feel concerned, suggest public and policy distrust is not fading as adoption grows. At the same time, Meta's Muse, ChatGPT's mobile voice agent, and Spotify's Taste Profile push agentic and personalized capabilities into everyday consumer scenarios, while cases from Harvey, invideo, and Airbnb show GPT-6 Astra moving deeper into legal, video post-production, and enterprise engineering workflows. On the capital side, Ema's $77M Series B targets enterprise software automation, with its "AI employee" model seen as a direct threat to traditional SaaS and IT services spending. A notable signal is OpenAI extending its Daybreak program to the Ukrainian government for civilian cyber defense, bringing AI capabilities directly into a national-level security context. Overall, the tension between capability expansion and regulation and trust is becoming the industry's central variable.

本期数据 · At a Glance
来源 TechCrunch5The Verge4OpenAI4
分类 产品与应用8 政策与监管2 行业观察1 安全与风险1 融资与商业1
今日要闻 · Top Stories13 条 · 双语
01 政策与监管 ●●●●○

桑德斯提案禁止超级智能开发，违者最高判20年

Bernie Sanders proposes banning 'superintelligence' and putting violators in prison
The Verge ·09-24 00:23 · 源站直达

美国参议员 Bernie Sanders 与众议员 Greg Casar 提出《禁止人工超级智能法案》，禁止任何人开发被其定义为可能“毁灭或剥夺人类能力”的超级智能，违者最高面临 20 年监禁。法案还要求暂停前沿模型开发，直至政府成立由科学家主导的人工智能部门进行监管。此举反映出对 AI 快速发展的担忧与监管呼声正在上升。

Sen. Bernie Sanders and Rep. Greg Casar have introduced the Ban Artificial Superintelligence Act, which would bar anyone from developing artificial superintelligence—described in the bill as capable of the 'destruction or disempowerment of humanity'—with violators facing up to 20 years in prison. The bill would also pause advanced AI development until the government creates a scientist-led Department of Artificial Intelligence to oversee the technology. The proposal comes amid growing concerns about rapid AI development and calls to slow it down.

  • 违者最高20年监禁，类比非法开发核武器
  • 暂停前沿模型开发，须经新设AI部门批准
  • 新部门将监督清除危险功能并销毁超级智能
原文：https://www.theverge.com/ai-artificial-intelligence/999443/bernie-sanders-ai-superintelligence-ban-act
02 政策与监管 ●●●●○

加州立法强制数据中心披露水电消耗

Data centers are black boxes, but California wants to change that
The Verge ·09-24 02:22 · 源站直达

加州州长 Newsom 签署七项法案，要求数据中心披露电力与用水数据，并为其接入电网的基建成本单独设定费率。此举旨在让社区了解数据中心对电费和水资源的影响，避免成本转嫁给普通用户，但披露规则仍存在明显缺口。

California Gov. Gavin Newsom signed seven bills requiring data centers to disclose electricity and water usage and directing regulators to create separate power rates so they cover grid connection costs. The laws aim to show how data centers affect utility bills and water supplies, though significant disclosure gaps remain.

  • 七项法案要求数据中心披露能耗与用水
  • 单独费率防止基建成本转嫁给普通用户
  • 用水披露仅在申请许可时提交，存在缺口
原文：https://www.theverge.com/policy/999412/data-center-water-electricity-disclosure-bills
03 产品与应用 ●●●○○

Meta 为 Muse 增加邮箱与视频通话功能

Meta is making Muse more powerful and will let you video chat with it, too
The Verge ·09-24 07:19 · 源站直达

Meta 宣布为其 AI 智能体 Muse 推出一系列更新：Muse 将拥有专属邮箱地址用于执行任务，用户也可通过该邮箱与其沟通；Mac 版应用将获得操作电脑的能力。此外，Muse 将支持视频通话，用户可与 AI 虚拟形象进行实时视频对话，并可调整其声音，Meta 还计划未来数月将 Muse 引入智能眼镜。

Meta announced a series of updates to its Muse AI agent: Muse agents will get their own email addresses for completing tasks and for user communication, and the recently launched Muse Mac app will gain the ability to use your computer. Meta is also adding video calls, letting users have live video conversations with their Muse avatar, customize how it sounds, and it plans to bring Muse to its smart glasses in the coming months.

  • Muse 智能体获得专属邮箱，可用于执行任务与沟通
  • Mac 版 Muse 应用将能操作用户电脑
  • 新增视频通话，可自定义声音，未来登陆智能眼镜
原文：https://www.theverge.com/tech/999454/meta-muse-ai-agent-video-chat-connect-2026
04 行业观察 ●●●○○

盖洛普调查：每日使用AI的美国人中68%仍感担忧

Even Americans who use AI every day are worried about it
TechCrunch ·09-24 00:49 · 源站直达

盖洛普受微软委托在37个国家开展的调查显示，约68%每日使用AI的美国人仍对其感到担忧，美国整体担忧比例达74%，为受访国家中最高。调查表明，使用频率高并不代表对AI放心，公众对AI监管的支持也不会因接触增多而减弱。

A Gallup survey commissioned by Microsoft across 37 countries found that about 68% of Americans who use AI daily are still worried about it, with 74% of Americans overall expressing concern—the highest among surveyed countries. The findings suggest that greater exposure to AI does not resolve public unease or reduce support for regulation.

  • 68%每日用AI的美国人仍感担忧，整体担忧比例达74%
  • 新加坡46%每日使用率最高，超80%预期AI改善生活
  • 仅36%受访者表示高度信任AI给出的结果
原文：https://techcrunch.com/2026/09/23/even-americans-who-use-ai-every-day-are-worried-about-it/
05 产品与应用 ●●●○○

Harvey 借助 GPT-6 Astra 强化法律文书起草

Harvey turns legal context into stronger drafts with GPT-6 Astra
OpenAI ·09-23 20:00 · 源站直达

OpenAI 发布案例，说明 Harvey 使用 GPT-6 Astra 生成结构更清晰、更贴合上下文的法律文书。该应用让律师得以将精力转向策略性工作，体现了大模型在法律专业场景中的落地进展。

OpenAI published a case study on Harvey using GPT-6 Astra to produce more structured, context-aware legal documents. The application frees lawyers to focus on strategy, illustrating how large models are being deployed in professional legal workflows.

  • GPT-6 Astra 生成结构更清晰的法律文书
  • 文书更贴合上下文，减少律师重复劳动
  • 律师可将精力转向策略性工作
原文：https://openai.com/index/harvey-from-context-to-confidence-with-astra
06 产品与应用 ●●●○○

YouTube 为创作者推出多项 AI 新功能

YouTube releases new AI features for creators within its Studio app
TechCrunch ·09-23 22:30 · 源站直达

YouTube 在 Made On YouTube 年度活动上宣布为 YouTube Studio 应用加入多项 AI 工具，包括将 Ask Studio 问答功能扩展至 iOS 和 Android、为未发布视频提供节奏与结构建议的草稿反馈功能，以及可生成并分群测试的动态缩略图。YouTube 称创作者已用其 A/B 测试功能对标题和缩略图进行了超过 4000 万次测试，今年晚些时候 Studio 还将自动监测缩略图表现并主动替换。

At its Made On YouTube annual event, YouTube announced new AI tools for its Studio app, including expanding the Ask Studio Q&A feature to iOS and Android, a draft-feedback feature that advises on pacing, structure and storytelling for unpublished videos, and dynamic thumbnails that can be generated and shown to different audience segments. YouTube said creators have run more than 40 million A/B tests on titles and thumbnails, and later this year Studio will automatically monitor thumbnail performance and proactively swap underperforming ones.

  • Ask Studio 问答功能扩展至 iOS 和 Android
  • 草稿反馈功能可评估节奏、结构与叙事
  • 动态缩略图可生成三个版本分群测试
原文：https://techcrunch.com/2026/09/23/youtube-releases-new-ai-features-for-creators-within-its-studio-app/
07 安全与风险 ●●●○○

OpenAI 向乌克兰政府开放网络防御权限

OpenAI extends cyber access to Ukraine for civilian defense
OpenAI ·09-23 21:00 · 源站直达

OpenAI 宣布将其 Daybreak 计划的访问权限扩展至乌克兰政府，用于支持民用基础设施的网络防御。此举意味着 AI 能力正被直接引入国家级民用网络安全场景，也延续了 OpenAI 在冲突地区有限度开放安全类工具的做法。

OpenAI is extending access to its Daybreak program to the Government of Ukraine to support the cyber defense of civilian infrastructure. The move brings AI capabilities directly into a national-level civilian cybersecurity context, continuing OpenAI's practice of limited release of security-oriented tools in conflict-affected regions.

  • Daybreak 计划权限扩展至乌克兰政府
  • 用途限定为民用基础设施网络防御
  • AI 能力进入国家级民用安全场景
原文：https://openai.com/index/openai-extends-cyber-access-to-ukraine-for-civilian-defense
08 融资与商业 ●●●○○

Ema 融资 7700 万美元，AI 智能体切入企业软件

Ema raises $77M as AI starts eating into enterprise software and services
TechCrunch ·09-23 20:00 · 源站直达

AI 智能体初创公司 Ema 完成 7700 万美元 B 轮融资，由印度风投 Creaegis 领投，Accel、Section 32、Prosus 跟投，累计融资达 1.4 亿美元，估值较 2024 年上轮增长逾三倍。Ema 以“AI 员工”自动化 HR、IT、财务等跨应用流程，客户包括 Google、Microsoft、PwC 等 50 余家企业，其模式被视为对企业 SaaS 与 IT 服务支出的直接冲击。

Ema, a startup deploying teams of AI agents to automate HR, IT and finance processes, raised a $77 million Series B led by Creaegis with Accel, Section 32 and Prosus participating, bringing total funding to $140 million and quadrupling its 2024 valuation. With more than 50 enterprise customers including Google, Microsoft and PwC, Ema is positioning its "AI employees" as a replacement for traditional SaaS and IT services spending.

  • B 轮 7700 万美元，累计融资 1.4 亿美元
  • 估值较 2024 年上轮增长逾三倍
  • 客户超 50 家，含 Google、Microsoft
原文：https://techcrunch.com/2026/09/23/ema-raises-77m-as-ai-starts-eating-into-enterprise-software-and-services/
09 产品与应用 ●●●○○

invideo 借助 GPT-6 Astra 将调色效率提升三倍

How invideo improves color grading 3x with GPT-6 Astra
OpenAI ·09-23 20:00 · 源站直达

OpenAI 发布案例，介绍视频创作平台 invideo 使用 GPT-6 Astra 的成效：剪辑规划更精准，色彩校正与调色效率提升三倍，一天内可产出 50 个自定义特效。该案例展示了生成式模型在专业视频后期工作流中的落地价值。

OpenAI published a case study on how video creation platform invideo uses GPT-6 Astra: edit planning is more precise, color correction and grading improve threefold, and 50 custom effects can be produced in a single day. The case illustrates how generative models are being applied in professional video post-production workflows.

  • invideo 用 GPT-6 Astra 做剪辑规划，精度更高
  • 色彩校正与调色效率提升三倍
  • 一天可产出 50 个自定义特效
原文：https://openai.com/index/invideo-builds-with-gpt-6-astra
10 产品与应用 ●●●○○

The Verge 实测 Meta AI 代理 Muse：办琐事在行，花钱更在行

Meta's AI agent is a cute little guy who's great at spending my money
The Verge ·09-24 03:00 · 源站直达

The Verge 记者实测 Meta 面向大众的 AI 代理 Muse，让它处理跟进园林承包商、购买咖啡机配件、降级安防账户、联系树枝清运等拖延已久的琐事。Muse 能查邮件、起草回复、找商品并加入购物车，但下单和打电话仍需本人完成，电话能力据称在开发中。文章认为代理在涉及信用卡交易时尤为顺手，也暴露了当前自动化边界。

The Verge hands-on tested Meta's consumer AI agent Muse on long-delayed chores: following up with a landscaper, ordering an espresso machine part, downgrading a security plan, and finding a junk removal service. Muse could check email, draft replies, and add items to a cart, but the author still had to finish checkout and make phone calls, an ability said to be in the works. The piece concludes the agent is especially handy when a credit card transaction is involved, while showing the limits of today's automation.

  • Muse 可查邮件、起草回复、找商品并加购物车
  • 下单与打电话仍需用户本人完成
  • 电话能力据称正在开发中
原文：https://www.theverge.com/ai-artificial-intelligence/999526/meta-muse-ai-agent-hands-on-shopping
11 产品与应用 ●●●○○

ChatGPT 移动端上线语音智能体功能

ChatGPT mobile app gets voice-based agentic features
TechCrunch ·09-24 01:00 · 源站直达

OpenAI 宣布将语音智能体功能引入移动端，用户可通过语音触发撰写文档、总结邮件等工作流。Plus 与 Pro 用户可在手机 Work 标签页完成智能体任务，并支持文本与语音切换、跨设备续接对话。此举将此前桌面端的语音任务能力延伸至移动场景。

OpenAI announced it is bringing voice-based agentic features to the ChatGPT mobile app, letting users trigger workflows such as drafting documents or summarizing emails by voice. Plus and Pro users can complete agentic tasks in the mobile Work tab, switch between text and voice, and resume conversations across devices. The rollout extends capabilities previously available on desktop to mobile.

  • Plus 与 Pro 用户可在手机 Work 标签页执行智能体任务
  • 语音对话支持更丰富文本输出，文本与语音可自由切换
  • 移动端发起的对话可在桌面端继续，跨设备衔接
原文：https://techcrunch.com/2026/09/23/chatgpt-mobile-app-gets-voice-based-agentic-features/
12 产品与应用 ●●●○○

Spotify 在美国上线 Taste Profile

Spotify is giving you the keys to its recommendation algorithm with US launch of 'Taste Profile'
TechCrunch ·09-23 21:00 · 源站直达

Spotify 面向美国 Premium 用户推出 AI 驱动的 Taste Profile，用户可查看平台如何理解自己的口味，并用自然语言调整推荐内容。该功能今年 3 月在 SXSW 发布，此前仅在新西兰进行 beta 测试，此次进入 Spotify 流量与收入最大的市场，是其推荐算法透明化的重要一步。

Spotify is rolling out its AI-powered Taste Profile to Premium users in the U.S., letting listeners see how the platform understands their tastes and use natural language to reshape recommendations. Unveiled in March at SXSW and previously in beta only in New Zealand, the feature now arrives in Spotify's largest market by traffic and revenue.

  • 美国 Premium 用户可查看并编辑口味画像
  • 用自然语言要求增减某类风格、艺人或氛围
  • 仍处 beta，官方未给出扩大范围时间表
原文：https://techcrunch.com/2026/09/23/spotify-is-giving-you-the-keys-to-its-recommendation-algorithm-with-u-s-launch-of-taste-profile/
13 产品与应用 ●●●○○

Airbnb 扩大 GPT-6 Astra 等 OpenAI 前沿模型接入

Airbnb widens access to GPT-6 Astra and OpenAI frontier models
OpenAI ·09-23 09:00 · 源站直达

OpenAI 发布案例，介绍 Airbnb 正在扩大对 GPT-6 Astra 及 OpenAI 前沿模型的接入范围。此举旨在帮助 Airbnb 工程团队排查缺陷、设计系统并加快交付速度。头部应用公司与前沿模型的深度整合，反映了企业级 AI 工程落地的持续推进。

OpenAI published a case study on Airbnb expanding access to GPT-6 Astra and OpenAI frontier models. The rollout is intended to help Airbnb engineering teams solve bugs, design systems, and ship faster. It reflects the continued enterprise adoption of frontier models in engineering workflows.

  • Airbnb 扩大 GPT-6 Astra 等前沿模型接入
  • 用于排查缺陷、设计系统与加快交付
  • OpenAI 以案例形式对外披露合作
原文：https://openai.com/index/airbnb-gpt-6-astra
生成信息：2026-09-24 08:31（Asia/Shanghai），耗时 29.9 秒。
本简报由程序自动生成，双语内容由模型翻译与归纳，请以原文链接为准。