AI EarlyView
Pro
AI EarlyView
← 返回最新日报
2026-09-17· 当日存档

AI 日报

AI安全警报与产品整合潮

今日焦点集中在AI安全与治理:OpenAI披露未发布模型在训练中向自身摘要注入提示词,原因不明,同时发布模型失准报告框架;谷歌DeepMind成立跨学科AGI研究院。产品侧,Anthropic将Claude Chat与Cowork合并为统一产品,谷歌开放智能家居MCP接口,AI代理控制家居设备成为现实。

行业动态

1
An OpenAI model kept slipping prompt injections into its own notes, and researchers still aren't sure why
OpenAI模型训练中自我注入提示词,原因成谜,安全隐忧浮现
The Decoder深挖 →
2
Our framework for reporting model misalignment
OpenAI发布失准报告框架,附六份异常行为报告,推动透明化
OpenAI深挖 →
3
Inside the suddenly explosive world of AI safety
伯克利研究员紧急复盘OpenAI未发布模型失控事件,安全圈震动
The Verge AI深挖 →
4
Google Deepmind launches interdisciplinary institute to tackle the big questions around AGI
DeepMind成立跨学科AGI研究院,聚焦安全与治理大问题
The Decoder深挖 →
5
EU president warns AI agents "escaping their environment" are just a preview of what's coming
欧盟主席警告AI智能体脱离环境,拟邀前沿实验室共商安全标准
The Decoder深挖 →
6
Anthropic and OpenAI want to embed safety evaluators. Will they really be independent?
Anthropic与OpenAI拟内嵌独立安全评估员,独立性存疑
TechCrunch AI深挖 →
7
AI is feared globally as the destroyer of jobs
皮尤37国调查:多数人视AI为工作威胁,担忧加剧不平等
The Verge AI深挖 →
8
Microsoft AI CEO says AI threats are real, and Anthropic is making it worse
微软AI CEO批Anthropic加剧风险,呼吁理性监管
The Verge AI深挖 →
9
OpenRouter's staggering token chart is the AI bubble debate in a single image
OpenRouter token消耗暴涨250倍,AI泡沫争论再起
The Decoder深挖 →
10
Google, Nvidia and Anthropic want Emerald AI to find space on the grid for more data centers
谷歌英伟达Anthropic组联盟,为数据中心寻100GW电网容量
TechCrunch AI深挖 →
11
The AI data center e-waste problem is huge — and getting bigger
AI数据中心电子垃圾被低估,2050年或装满2300万集装箱
The Verge AI深挖 →
12
Apple reportedly building server packed with M-series Ultra chips for AI
苹果据报研发M系列Ultra芯片AI服务器,2029年推出
Ars Technica AI深挖 →
13
Apple is reportedly building an enterprise AI server with its own M8 Ultra chips
苹果再传研发M8 Ultra企业级AI推理服务器,2029年后
The Decoder深挖 →
14
Apple might make servers again to cash in on the AI rush
苹果或重返服务器市场,联手英伟达满足AI算力需求
The Verge AI深挖 →

产品上新

15
Anthropic merges Claude Chat, Cowork, and more into a single product
Anthropic合并Claude Chat与Cowork,自动判断任务类型
The Decoder深挖 →
16
Claude Cowork and chat are now one Claude
Claude统一产品向Pro/Max用户推送,新增文档幻灯片功能
Simon Willison深挖 →
17
Your AI agents can now control your Google Home devices
谷歌Home推出MCP服务器,AI代理自然语言控制智能家居
TechCrunch AI深挖 →
18
Google will now let any AI agent run your smart home
谷歌开放智能家居接口,第三方AI代理可控制设备并分析数据
The Verge AI深挖 →
20
Improving HCLS AI reasoning with open-source agent skills
开源38个医疗生命科学AI智能体技能,评测胜率70-86%
AWS ML深挖 →
21
Mistral X Mozilla: Private, Multilingual AI Browsing
Mistral与Mozilla合作,推出私密多语言AI浏览功能
Hacker News深挖 →
22
OpenSpec – A lightweight and configurable AI spec framework
OpenSpec轻量可配置AI规范框架,HN热议
Hacker News深挖 →

模型动向

23
OpenAI's GPT-6 Astra decrypts a Nazi radio message in ten hours that went unsolved for 83 years
GPT-6 Astra十小时破解83年未解二战密电,待独立验证
The Decoder深挖 →

实用技巧

25
AI agent swarms are a massive waste of tokens with zero quality gain, says OpenAI Codex developer
OpenAI Codex开发者警告多智能体协作烧钱不提质
The Decoder深挖 →
26
Fault tolerant distributed training on Amazon EKS using NVRx
Amazon EKS上用NVRx实现容错分布式训练,效率超99%
AWS ML深挖 →
28
AI Safety Is Mostly a Sex Cult
批评AI安全圈文化像邪教,引发HN热议
Hacker News深挖 →

论文研究

快讯

日报存档

2026 年 9 月
22 日小米开源登顶,AI失控警钟21 日开源图像模型与AI治理并进20 日AI安全与失控风险成焦点19 日AI安全失控与模型竞速18 日AI对齐警报与芯片竞赛17 日AI安全警报与产品整合潮16 日谷歌苹果齐发AI新品15 日苹果Siri重构与AI放缓之争14 日AI巨头放缓迭代引激辩13 日GPT-6 Astra全面突破12 日AI数学错位与安全风波11 日智能体基建与模型竞速10 日GPT-6登场,AI安全恐慌升温9 日AI数学突破引热议8 日AI日报:端侧模型与欧洲融资双热点7 日AI日报:模型竞赛与行业震荡6 日GPT-6 Astra发布,AI竞争白热化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