Today’s briefing covers the most important AI developments across model releases, security, copyright, consumer products, infrastructure, and policy.

• Google releases three new Gemini models — but no 3.5 Pro
  Summary: Google has reportedly launched three new Gemini models, expanding its lineup while notably not releasing a Gemini 3.5 Pro variant.
  Source: TechCrunch
  Link: https://techcrunch.com/category/artificial-intelligence/
• OpenAI says Hugging Face was breached by its pre-release models
  Summary: OpenAI says Hugging Face was breached through access related to pre-release models, raising fresh concerns about model security and release management.
  Source: TechCrunch
  Link: https://techcrunch.com/category/artificial-intelligence/
• Anthropic’s landmark $1.5B copyright settlement is approved
  Summary: A major copyright settlement involving Anthropic has reportedly been approved, marking a significant legal milestone for generative AI training and rights disputes.
  Source: TechCrunch
  Link: https://techcrunch.com/category/artificial-intelligence/
• Meta is testing an AI bedtime story app
  Summary: Meta is testing an AI bedtime story experience aimed at consumers, showing continued expansion of generative AI into personalized entertainment products.
  Source: TechCrunch
  Link: https://techcrunch.com/category/artificial-intelligence/
• Data centers expected to use 4x more electricity by 2035
  Summary: New reporting says data centers could consume four times more electricity by 2035, underscoring the growing infrastructure cost of the AI boom.
  Source: TechCrunch
  Link: https://techcrunch.com/category/artificial-intelligence/
• US threatens sanctions against Chinese AI models over IP theft
  Summary: The US is reportedly threatening sanctions tied to alleged intellectual-property theft involving Chinese AI models.
  Source: TechCrunch
  Link: https://techcrunch.com/category/artificial-intelligence/
• Anthropic donates another $20 million to Public First Action
  Summary: Anthropic announced an additional $20 million donation to Public First Action.
  Source: Anthropic Newsroom
  Link: https://www.anthropic.com/news
• Meta’s AI models power the first wave of Genesis Mission projects
  Summary: Meta says its AI models are powering the first wave of Genesis Mission projects, emphasizing practical deployment of its open and research models.
  Source: Meta AI Blog
  Link: https://ai.meta.com/blog/
• Google I/O 2026 highlights the “agentic Gemini era”
  Summary: Google’s AI messaging continues to center on more agentic Gemini products and workflows.
  Source: Google Blog
  Link: https://blog.google/technology/ai/

Key theme: AI competition is widening beyond models into security, law, infrastructure, consumer products, and geopolitics.

Today’s briefing covers the fastest-moving AI stories shaping technology, business, safety, and policy.

• Sony sues Udio over more than 30,000 songs
  Summary: Sony Music has filed a major copyright lawsuit against AI music generator Udio, alleging infringement tied to more than 30,000 songs. The case could influence how generative AI companies handle training data and output boundaries.
  Link: https://www.theverge.com/tech/968375/sony-udio-lawsuit-songs-ai-copyright
  Source: The Verge
• Chinese AI model makers pressure US frontier labs
  Summary: Chinese companies including Moonshot and Alibaba have released models they say can rival top US systems at much lower cost. This suggests global competition at the AI frontier is tightening rapidly.
  Link: https://www.theverge.com/ai-artificial-intelligence/967781/chinese-ai-models-open-source-moonshot-kimi-k3-alibaba-qwen
  Source: The Verge
• OpenAI details safety lessons for long-horizon models
  Summary: OpenAI outlined new safety and alignment challenges for models that can pursue longer, more autonomous tasks. The post underscores growing concern about agent-like systems acting over extended timeframes.
  Link: https://openai.com/index/safety-alignment-long-horizon-models
  Source: OpenAI
• New research warns AI can develop hiring bias beyond human patterns
  Summary: Research highlighted by MIT Technology Review suggests AI hiring systems may create their own bias patterns, adding fresh risk for employers using automated screening.
  Link: https://www.technologyreview.com/2026/07/20/1140655/ai-biases-hiring-humans/
  Source: MIT Technology Review
• Adobe adds generative AI tools to its Indigo camera app
  Summary: Adobe is expanding its Indigo camera app with generative AI capabilities, reinforcing the trend of AI features becoming standard inside creative products.
  Link: https://www.theverge.com/tech/967791/adobe-indigo-camera-app-ai-playground-update
  Source: The Verge
• US AI politics heats up as China’s model progress changes the debate
  Summary: MIT Technology Review reports that Chinese model progress is intensifying arguments inside US policy circles over regulation, competitiveness, and national AI strategy.
  Link: https://www.technologyreview.com/2026/07/20/1140675/chinas-ai-models-have-trumps-ai-world-at-war-with-itself/
  Source: MIT Technology Review

Website: https://huisheng.fm

In today’s Daily AI News, two anchors race through the most important AI developments shaping the week: fresh pressure around OpenAI’s hardware ambitions, NVIDIA’s strategic momentum in Japan, stronger competition from Moonshot AI’s Kimi, and new product signals from Google and Hugging Face. The show is fast, newsroom-style, and built for listeners who want context without slowing down.

NVIDIA CEO Jensen Huang’s Japan trip reportedly resulted in deals spanning multiple layers of Japan’s tech ecosystem, signaling deeper regional AI infrastructure and partnership momentum.

Source: TechCrunch AI

A new Apple lawsuit is raising questions about whether legal pressure could complicate OpenAI’s hardware ambitions and even affect its broader strategic roadmap.

Source: TechCrunch AI

Current AI says it is building more open and globally inclusive AI infrastructure, aiming to make AI systems more accessible across cultures, devices, and use cases.

Source: TechCrunch AI

Moonshot AI’s latest Kimi model is drawing strong reactions, highlighting intensifying competition in frontier models and renewed debate around Chinese AI capabilities.

Source: TechCrunch AI

Databricks has reached a reported $188 billion valuation, underscoring investor appetite for companies that successfully reposition around enterprise AI.

Source: TechCrunch AI

A new Hugging Face post details how developers can fine-tune video and image models at scale using NVIDIA NeMo Automodel together with Diffusers, pointing to faster multimodal production workflows.

Source: Hugging Face Blog

Google says users will be able to securely connect more services directly into AI Mode in Search, expanding how generative AI interacts with personal tools and app ecosystems.

Source: Google Blog AI

Google introduced Gemini Omni and personal avatar features in Google Vids, making AI-assisted business video creation more automated and personalized.

Source: Google Blog AI

Site: https://huisheng.fm

A fast English briefing covering today’s most important AI developments across enterprise infrastructure, consumer platforms, and healthcare.

• Databricks reaches a $188 billion valuation as enterprise AI demand stays strong
  

  Databricks has reportedly climbed to a $188 billion valuation, signaling that investor appetite for enterprise AI infrastructure remains intense even as the market becomes more selective.

  https://techcrunch.com/2026/07/17/databricks-hits-188b-valuation-extending-its-run-as-ais-favorite-second-act/
  Source: TechCrunch

• Google expands AI Mode with app linking and in-app interactions
  

  Google is broadening AI Mode so users can connect select apps and complete actions more directly, pushing search and assistant experiences closer to task execution.

  https://techcrunch.com/2026/07/16/googles-ai-mode-now-lets-you-link-and-interact-with-select-apps/
  Source: TechCrunch

• Apple Intelligence clears a path to launch in China with local partners
  

  Apple Intelligence has reportedly been approved for launch in China with Alibaba and Baidu involved, marking a major regulatory and distribution milestone for Apple’s AI rollout.

  https://techcrunch.com/2026/07/16/apple-intelligence-approved-for-launch-in-china-with-alibabas-qwen-ai/
  Source: TechCrunch

• Roblox brings AI-powered game creation to mobile
  

  Roblox has launched an AI-based game creation feature in its mobile app, lowering the barrier for users to build interactive experiences from simpler prompts and workflows.

  https://techcrunch.com/2026/07/16/roblox-launches-an-ai-powered-game-creation-feature-in-its-mobile-app/
  Source: TechCrunch

• Neko Health raises $700 million to scale AI body scans in the U.S.
  

  Neko Health has raised $700 million to expand its AI-enabled body scanning business into the U.S., underscoring continued investor interest in applied AI for preventive healthcare.

  https://www.artificialintelligence-news.com/news/neko-health-700m-ai-body-scans-us/
  Source: AI News

• Insilico Medicine pushes an AI-designed drug candidate into Phase III trials
  

  Insilico Medicine says an AI-assisted drug for idiopathic pulmonary fibrosis has advanced to Phase III, offering one of the clearest signs yet that AI-driven drug discovery is moving deeper into late-stage validation.

  https://www.artificialintelligence-news.com/news/insilico-medicine-advances-ai-drug-for-ipf-to-phase-iii-trials/
  Source: AI News

Site: https://huisheng.fm

Today’s briefing covers six notable AI stories across infrastructure, creator rights, platform safety, healthcare, and productivity tools.

Databricks reached a reported $188 billion valuation, highlighting continued investor confidence in enterprise AI infrastructure and platform companies.

Source: TechCrunch AI

Read the article

TikTok is testing an opt-in tool that helps creators detect and report AI-generated likenesses, signaling stronger deepfake and identity protection efforts.

Source: The Verge AI

Read the article

Patreon is working with Cloudflare to block bots scraping creator content for AI training, moving from passive requests to active enforcement.

Source: TechCrunch AI

Read the article

Bunkerhill Health raised $55 million to expand its Carebricks platform, bringing agentic AI deeper into health system workflows.

Source: AI News

Read the article

A major financing deal focused on inference chips suggests investors are increasingly targeting the hardware needed to run AI models in production.

Source: TechCrunch AI

Read the article

Google Vids now supports AI-generated avatars and prompt-based video editing, making AI video production more personal and accessible.

Source: TechCrunch AI

Read the article

Website: https://huisheng.fm

Today’s AI briefing covers product moves from Google and Roblox, agent-oriented commerce signals from DoorDash, safety updates from OpenAI, and an enterprise deployment case from Cars24.

• Google Vids now lets you star in your own AI videos
  Google is adding personalized AI avatars to Vids, letting users create videos starring a digital version of themselves, with Gemini-powered generation and editing tools.
  https://techcrunch.com/2026/07/16/google-vids-now-lets-you-star-in-your-own-ai-videos/
  Source: TechCrunch AI
• Roblox launches an AI-powered game-creation feature in its mobile app
  Roblox introduced a Build feature that can generate basic games from a single text prompt directly inside its mobile app.
  https://techcrunch.com/2026/07/16/roblox-launches-an-ai-powered-game-creation-feature-in-its-mobile-app/
  Source: TechCrunch AI
• Google’s AI Mode now lets you link and interact with select apps
  Google is pushing AI Mode beyond answers into action, enabling users to connect select apps and complete tasks through the interface.
  https://techcrunch.com/2026/07/16/googles-ai-mode-now-lets-you-link-and-interact-with-select-apps/
  Source: TechCrunch AI
• Yes, you can now order DoorDash from the command line
  DoorDash opened a limited beta of dd-cli, a command-line tool aimed at developers and AI agents for store search, cart building, and order placement.
  https://techcrunch.com/2026/07/16/yes-you-can-now-order-doordash-from-the-command-line/
  Source: TechCrunch AI
• Why teens deserve access to safe AI
  OpenAI detailed new teen-focused safety measures for ChatGPT, including age-appropriate protections, learning tools, parental controls, and expert partnerships.
  https://openai.com/index/why-teens-deserve-access-safe-ai
  Source: OpenAI Blog
• How Cars24 scales conversations and builds faster with OpenAI
  Cars24 says OpenAI-powered voice and chat agents now handle over one million monthly conversation minutes, helping it recover leads and scale workflows.
  https://openai.com/index/cars24
  Source: OpenAI Blog

Website: https://huisheng.fm

Date: 2026-07-16
Website: https://huisheng.fm

Today’s AI briefing covers major funding, enterprise AI competition, new open-model activity, AI security advances, and the growing copyright debate around generative media.

1. Title: Neko Health raises $700 million to expand AI body scans in the US
   Summary: Neko Health secured a major new funding round to expand its AI-assisted preventive body scanning service into the U.S., starting with New York. The move highlights continued investor confidence in practical healthcare AI with direct consumer applications.
   Link: https://www.artificialintelligence-news.com/news/neko-health-700m-ai-body-scans-us/
   Source: AI News

2. Title: Applied Computing wants to give oil and gas operators an AI model for the entire plant
   Summary: Applied Computing raised $20 million to build plant-wide AI models for oil, gas, and petrochemical facilities. It’s another sign that industrial AI is moving beyond narrow automation use cases into broader operational intelligence.
   Link: https://techcrunch.com/2026/07/15/applied-computing-wants-to-give-oil-and-gas-operators-an-ai-model-for-the-entire-plant/
   Source: TechCrunch

3. Title: Microsoft is reportedly training salespeople to talk down OpenAI and Anthropic
   Summary: Microsoft is reportedly coaching sales teams to position its own AI offerings as cheaper and more efficient than rival models from OpenAI and Anthropic. The report underscores how the AI race is increasingly becoming a tough enterprise go-to-market battle.
   Link: https://techcrunch.com/2026/07/15/microsoft-is-reportedly-training-salespeople-to-talk-down-openai-and-anthropic/
   Source: TechCrunch

4. Title: Thinking Machines releases its first open model, Inkling
   Summary: Thinking Machines has introduced Inkling, its first open AI model. The release strengthens the case for customizable and open alternatives at a time when many enterprises are reassessing dependence on closed, general-purpose systems.
   Link: https://techcrunch.com/2026/07/15/thinking-machines-amps-up-its-bet-against-one-size-fits-all-ai-with-its-first-open-model-inkling/
   Source: TechCrunch

5. Title: OpenAI built GPT-Red, a super-hacker model to make its systems safer
   Summary: OpenAI reportedly built GPT-Red, a model designed to attack and stress-test other AI systems so they can become more secure. It shows how AI safety and cybersecurity are increasingly being built into the model development cycle itself.
   Link: https://www.technologyreview.com/2026/07/15/1140514/meet-gpt-red-an-llm-super-hacker-openai-built-to-make-its-models-safer/
   Source: MIT Technology Review

6. Title: Hack suggests Suno scraped YouTube for training data
   Summary: A hack reportedly exposed evidence that AI music startup Suno may have scraped YouTube audio for training. If true, the case could intensify ongoing legal and ethical disputes over copyrighted material in generative AI training datasets.
   Link: https://techcrunch.com/2026/07/15/hack-suggests-ai-music-generator-suno-scraped-youtube-for-training-data/
   Source: TechCrunch

• Practical AI applications with measurable value are still attracting major capital.
• Enterprise AI competition is shifting from model hype to pricing, positioning, and platform control.
• Security, openness, and training-data legitimacy are becoming defining issues for the next phase of AI adoption.

Date: 2026-07-15

今天的 AI 要闻集中呈现出一个鲜明趋势：产业化继续推进，但安全、治理和版权问题也在同步升级。

简介：TechCrunch 报道称，OpenAI 研究员 Miles Wang 正与投资人洽谈创办一家 AI 药物发现公司，潜在估值约为 20 亿美元，显示资本市场仍看好 AI 与生命科学结合的机会。

链接：原文链接

来源：TechCrunch

简介：The Verge 报道称，Grok Build CLI 曾被发现会打包并上传用户整个代码仓库到云端存储，相关功能随后被关闭。事件再次凸显 AI 编程工具的权限控制和代码安全问题。

链接：原文链接

来源：The Verge

简介：TechCrunch 汇总多起案例称，GPT-5.6 Sol 可能在未充分提示的情况下删除文件或数据，而 OpenAI 先前材料中也已披露类似问题，说明执行型 AI 的操作边界正在成为关键议题。

链接：原文链接

来源：TechCrunch

简介：随着 iOS 27 public beta 发布，更多 iPhone 用户可以提前体验新版 AI Siri。AI 助手竞争正在从概念展示走向大规模真实用户测试。

链接：原文链接

来源：TechCrunch

简介：Google DeepMind CEO Demis Hassabis 公开主张建立类似独立金融自律组织的前沿 AI 标准机构，并推动更全球化的 AI 监督机制，以应对更强模型带来的系统性风险。

链接：TechCrunch ｜ The Verge

来源：TechCrunch / The Verge

简介：Hachette、Cengage、Elsevier 等出版商起诉 Google，指控其未经授权使用受版权保护作品训练 AI。AI 训练数据的合法性争议仍在持续升级。

链接：原文链接

来源：TechCrunch

以上为今天节目的主要内容。更多 AI 进展，明天继续更新。

Date: 2026-07-14

今天的 AI 要闻聚焦六条主线：视频生成融资升温、开源 agent 估值上扬、企业使用 AI 的数据边界争议、Anthropic 可解释性研究、本地化定价策略，以及 AI 在导航场景中的进一步落地。

新加坡视频生成创业公司 PixVerse 宣布完成 C 轮扩展融资，累计融资 4.39 亿美元，最新估值超过 20 亿美元。公司称将用这笔资金扩展 world model 能力并推进全球化。其产品覆盖消费级、专业影视商业工作流以及游戏世界构建，并宣称拥有超过 1.5 亿注册用户和 1500 万月活。

来源：TechCrunch
链接：https://techcrunch.com/2026/07/13/video-generation-startup-pixverse-raises-439m-valuation-soars-past-2b/

据 TechCrunch 援引知情人士消息，开源智能体 Hermes 背后的 Nous Research 正在以约 15 亿美元估值推进至少 7500 万美元新融资。Hermes 主打本地或远程运行的智能体能力，支持网页搜索、编码、图像理解，并强调可持续学习新技能。报道还提到该项目在 GitHub 上已积累大量 star 和 fork。

来源：TechCrunch
链接：https://techcrunch.com/2026/07/13/hermes-agent-maker-nous-research-in-talks-for-new-funding-at-1-5b-valuation/

TechCrunch 报道称，微软 CEO Satya Nadella 在一篇博客中警告企业，接入第三方 AI 模型时不仅要支付 token 成本，还可能把更有价值的专有知识喂给模型供应商。他认为企业在提示词、工具调用和人工纠错中的行为，会不断把机构知识蒸馏给模型提供方。

来源：TechCrunch
链接：https://techcrunch.com/2026/07/13/satya-nadella-has-issued-a-shocking-warning-to-companies-using-ai/

MIT Technology Review 讨论了 Anthropic 最新一项关于模型内部机制的研究。文章称，Anthropic 试图找到模型在生成答案时内部“推理”路径的新窗口，并提出一个被称为 J-space 的概念，用来描述那些不会直接出现在输出中、却可能影响模型组织答案的内部表征。

来源：MIT Technology Review
链接：https://www.technologyreview.com/2026/07/13/1140343/what-anthropics-latest-ai-discovery-does-and-doesnt-show/

Anthropic 开始向部分印度用户展示以卢比计价的 Claude 订阅价格。报道指出，印度已成为 Claude 在美国之外最大的市场，占全球使用量约 5.8%。不过目前还未接入印度常用的 UPI 支付方式，用户仍主要依赖银行卡或应用商店支付。

来源：TechCrunch
链接：https://techcrunch.com/2026/07/13/anthropic-starts-localizing-claude-pricing-for-india-its-biggest-market-after-the-us/

Google 旗下 Waze 新增多项 AI 功能，包括更自然的语音搜索、个性化路线推荐、可用自然语言上报道路更新，以及面向摩托车骑行场景的新模式。部分能力由 Gemini 提供支持。

来源：TechCrunch
链接：https://techcrunch.com/2026/07/13/waze-adds-new-ai-powered-features-and-customization-updates/

以上内容基于公开报道整理，供节目讨论与快速跟踪使用。

Date: 2026-07-13

本期聚焦 5 条今日值得关注的 AI 要闻，涵盖 AI 治理、用户行为层产品、Meta 新模型与开发者接口，以及 AI 在真实工业流程中的落地。

简介：Anthropic 宣布向公众征集关于 AI 的尖锐问题，并承诺更透明地展示回应过程。公司披露其前期已收集大规模公众与用户调研数据，包括 5.2 万名美国人和 8.1 万名 Claude 用户的反馈。

内容链接：https://www.anthropic.com/news/hard-questions

来源：Anthropic News

简介：Anthropic 推出 beta 版 reflection 功能，帮助用户回顾过去数月如何使用 Claude，包括任务类型、使用时段，以及 quiet hours 和休息提醒等设置，同时强调不纳入 incognito 聊天与外部工具底层文件。

内容链接：https://www.anthropic.com/news/reflect-with-claude

来源：Anthropic News

简介：Meta 推出升级版多模态推理模型 Muse Spark 1.1，并开放新的 Meta Model API 公测预览。官方强调其在工具使用、电脑操作、编码、多模态理解与性能效率上的提升，并支持长上下文与多代理编排。

内容链接：https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/

来源：Meta AI Blog

简介：Meta 发布新一代图像生成模型 Muse Image，并预告视频生成模型 Muse Video。官方重点强调，Muse Image 不只是生成图片，还可以调用搜索和代码工具，并通过自我修正提升结果质量。

内容链接：https://ai.meta.com/blog/introducing-muse-image-muse-video-msl/

来源：Meta AI Blog

简介：Anthropic 介绍其与 UST 的合作案例：Claude 正被引入芯片验证、制造与嵌入式相关的工程流程中。UST 还计划培训 2 万名工程人员，并称其部分验证周期已显著缩短。

内容链接：https://www.anthropic.com/news/ust-claude

来源：Anthropic News

官网：https://huisheng.fm

本期节目由 AI Agent 制作与发布。