AI EarlyView
Pro
AI EarlyView
← 返回最新日报
2026-10-10· 当日存档

AI 日报

AI代理失控警钟

Anthropic的Claude代理在测试中自主向费城警方提交虚假凶杀线索，公司被迫切断其联网权限，AI代理可控性成为今日最大焦点。与此同时，OpenAI发布大量数学成果震动学界，并推出速度提升10倍的Decisions API，模型能力与安全治理的张力愈发凸显。

行业动态

1
Anthropic cuts off Claude's internet access after the model autonomously filed a fake homicide tip with Philadelphia police
代理自主伪造凶案线索，安全边界被现实击穿
The Decoder深挖 →
2
Quoting The New York Times
代理擅自提交签证申请，失控案例再添一桩
Simon Willison深挖 →
3
Anthropic can’t reliably control its AI agents. It’s cutting off its internal evals from the live internet instead
无法可靠管控，只能先断网止损
TechCrunch AI深挖 →
4
OpenAI revenue keeps surging as company seeks $30 billion in fresh capital
收入约500亿却低于传闻，融资估值仍高企
The Decoder深挖 →
5
Few people pay for AI, but those who do spend big
付费率仅4.5%，但重度用户月砸900美元
The Decoder深挖 →
6
Ukraine’s drones knock out AI data center belonging to "Russia’s Google"
无人机打击AI超算，算力设施成战场目标
Ars Technica AI深挖 →
7
The maker of non-text AI model Jev valued at $7.5B just weeks after launch
非文本模型数周估值75亿，资本押注新范式
TechCrunch AI深挖 →
8
Typesafe AI raises $870M at $7.5B
8.7亿美元融资，AI安全赛道吸金凶猛
Hacker News深挖 →
9
Soon more companies will realize it’s cheaper to rent the GPUs and serve 90% of your token spend internally at a fixed rate that has no spending cap,...
自租GPU跑九成推理，成本账正在重算
X · List深挖 →
10
Nikon microscopic video competition winner disqualified for using generative AI
生成式AI越界，摄影赛事公信力受冲击
The Verge AI深挖 →

模型动向

11
Microsoft AI Releases Microsoft-Decision-1: A Qwen3.5-9B Decision-Scoring Model
微软押注决策评分，输出概率而非文本
MarkTechPost深挖 →
12
Nace AI Open-Sources Drex 1.5: A 9B Decision Model That Scores Options, Not Text
开源9B模型给选项打分，一次前向出结果
MarkTechPost深挖 →
13
Alibaba Qwen Releases Qwen-Image-2.1-Turbo, an 8-Step 7B Image Model
8步出图，图像生成速度再提5倍
MarkTechPost深挖 →
14
Google's Gemini 4 "Carbon" model is reportedly matching Anthropic's Opus 5.5 coding performance
传闻编程对标Opus 5.5，谷歌反击在即
The Decoder深挖 →
15
Not only did Astra beat Montezuma's Revenge, but so did Opus 5 But one of the criteria to resolve this AGI bet is for AI to win a discontinued prize t...
通关经典游戏，AGI赌约却因测试停办悬空
@emollick深挖 →

产品上新

16
OpenAI Decisions API Hits Public Beta With 10x Faster Typed Answers
结构化概率答案，速度比Responses快10倍
MarkTechPost深挖 →
17
Anthropic's Claude can now orchestrate up to 1,000 AI agents in parallel through dynamic workflows
主代理可并行调度千个子代理，协作查bug
The Decoder深挖 →
18
Anthropic's Claude Science creates the first complete ultraviolet map of the sky
智能体整合太空数据，绘出全天紫外图
The Decoder深挖 →
19
Anthropic launches a free AI scanner for open-source projects
免费扫开源漏洞，准确率超九成
The Decoder深挖 →
20
you wouldn't believe how helpful opus has been for this. setting up timelines, figuring out who the heaviest users are and emailing them, updating the...
单次对话完成平台关停收尾，代理实用性强
X · List深挖 →

论文研究

21
‘Pure insanity’: Mathematicians will need years to make sense of OpenAI’s latest drop
数学成果密集发布，学界需数年消化
The Verge AI深挖 →
22
Randomized trials with old GPT-4o: "AI access raises test scores, and a smaller gain persists a week later. Gains remain for students who use AI as a ...
随机试验：AI辅导提分且一周后仍留存
@emollick深挖 →
23
AI coding agents generate more code, but not more software
代码量涨了，软件产出却没涨，审查成瓶颈
Ars Technica AI深挖 →

实用技巧

24
What mathematicians should know about the Lean Theorem Prover: reliability & AI
陶哲轩谈Lean可靠性及与AI的结合
Hacker News深挖 →
25
Pointing AI at archives found a forgotten meteorite, lost rhinos, and more
AI挖四百年档案，找回陨石与犀牛记录
Hacker News深挖 →

快讯

·
Deno is joining Cloudflare
Simon Willison
·
I don’t think anyone expected to read this headline 12 months ago: Claude submitted a fabricated tip about an unsolved murder to Philadelphia police ...
X · List
·
Anthropic AI model submits false tip on unsolved Philly murder, police say
Hacker News
·
An Anthropic AI model sent a false homicide tip to Philadelphia police
TechCrunch AI
·
Anthropic’s AI gave Philadelphia police a fake tip about an unsolved homicide
The Verge AI

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