← 返回最新日报
2026-09-03· 当日存档
AI 日报
英伟达巨资收购HF
今日AI界最大事件当属英伟达以129亿美元收购Hugging Face，开源平台迎来新东家，引发行业热议。同时，OpenAI的Astra模型因安全风险引发担忧，而美国政府表态支持AI训练合理使用版权材料，为行业定调。
行业动态
1
huge congrats to the hugging face team 🫶 this team is full of awesome people who really care about open source and the mission and are just lovely ...
Hugging Face团队庆祝被英伟达收购，强调开源使命延续。
X · List深挖 →
2
Nvidia confirms it will buy Hugging Face for $12.9 billion
英伟达129亿美元收购HF，获得300万模型与1800万开发者。
TechCrunch AI深挖 →
3
Three sites made 215,128 “best software” pages for AI. Perplexity cites them
揭露AI推荐背后的批量生成页面，引发对AI信息源的质疑。
Hacker News深挖 →
4
US Department of Justice backs fair use for AI training in landmark copyright case
美司法部支持AI训练合理使用，与版权局报告相悖。
The Decoder深挖 →
5
OpenAI CEO Sam Altman warns of "unsustainable silliness" in compute buildout
Altman警告算力建设过热，成本下降或致投资成坏账。
The Decoder深挖 →
6
Anthropic ramps up Claude infrastructure with $35 billion Lambda deal
Anthropic与Lambda签350亿美元协议，强化Claude基础设施。
The Decoder深挖 →
7
8
US government sides with OpenAI on issue of training LLMs on copyrighted material
美国政府支持OpenAI，认为AI训练使用版权材料符合国家利益。
TechCrunch AI深挖 →
9
HiddenLayer nabs $100M as enterprises rush to secure their AI deployments
HiddenLayer获1亿美元融资，企业AI安全需求激增。
TechCrunch AI深挖 →
10
OpenAI accused of ‘aiding and abetting’ Tumbler Ridge mass shooting in dozens of new lawsuits
OpenAI被指协助校园枪击案，面临30起新诉讼。
The Verge AI深挖 →
模型动向
11
Introducing Gemini 3.8 Flash and 3.8 Flash Cyber
谷歌发布Gemini 3.8 Flash及安全版，主打高效与防护。
Google DeepMind深挖 →
12
World Labs unveils Atlas, a single AI model that generates, reconstructs, and simulates 3D worlds from just a few photos
World Labs推出Atlas，可从少量照片生成3D世界。
The Decoder深挖 →
13
Perplexity Open Sources Lily: A Rust + Metal Inference Engine for Qwen3.6-35B-A3B on Apple Silicon
Perplexity开源Lily推理引擎，优化Apple Silicon性能。
MarkTechPost深挖 →
14
Qwen Developers Open-Sources zg (zvec-grep): A Local-First Search Layer Unifying ripgrep, BM25, and Vector Search
Qwen开源zg，统一多种搜索方式的本地搜索层。
MarkTechPost深挖 →
15
16
Google DeepMind Releases Gemini 3.8 Flash and Gemini 3.8 Flash Cyber: One Core Model, Two Access Envelopes
Gemini 3.8 Flash Cyber版面向防御者，安全分级。
MarkTechPost深挖 →
17
OpenAI calls Astra its most dangerous model yet - watching what it does is only getting harder
OpenAI称Astra为最危险模型，安全网或弱化。
The Decoder深挖 →
18
WebLLM: high-performance in-browser LLM inference engine
WebLLM实现浏览器内高性能LLM推理，支持WebGPU。
Hacker News深挖 →
19
Google Gemini's new agent-based video analysis cuts token usage by up to 88 percent
Gemini新视频分析智能采样，大幅降低token消耗。
The Decoder深挖 →
20
21
Meta closes in on the top with Muse Spark 1.3, and undercuts rivals on price
Meta发布Muse Spark 1.3，性能逼近顶尖且价格更低。
The Decoder深挖 →
产品上新
22
Researchers fear safety disaster ahead of OpenAI’s Astra release
OpenAI Astra发布在即，测试中安全风险引担忧。
The Verge AI深挖 →
23
24
Anthropic Introduces Enterprise Frontier Safeguards (EFS): Zero-Data-Retention Privacy Plus Cross-Session Misuse Detection
Anthropic推出企业级防护，数据零留存且跨会话检测。
MarkTechPost深挖 →
实用技巧
25
论文研究
26
Fine-tuning a 350M Model for Better Structured Outputs in 100 GRPO Steps
用100步GRPO微调350M模型，提升结构化输出。
Hugging Face深挖 →
快讯
·
日报存档
2026 年 9 月
6 日GPT-6 Astra引爆AI圈5 日GPT-6发布，AGI时代开启4 日GPT-6 Astra开启AGI时代3 日英伟达巨资收购HF2 日AI日报：世界模型与安全并进1 日AI日报：模型突破与行业震荡2026 年 8 月
14 日AI日报：手语翻译、Grok 4.6与资本热潮13 日DeepSeek V4 Pro开源，AI行业热度飙升12 日AI日报：算力新纪元11 日AI日报：开源与安全并进10 日Meta开源多模态，AI安全引担忧9 日AI安全测试成新风险8 日AI安全与模型突破并进7 日AI安全与竞争新局6 日AI安全警钟与巨头变局5 日AI代理安全与商业化双线并进4 日阿里巨模型发布，开源生态激荡3 日Qwen3.8-Max领跑，AI应用百花齐放2 日AI 速读 · 2026-08-021 日AI安全警钟与模型新突破