AI EarlyView
Pro
AI EarlyView
← 返回最新日报
2026-08-05· 当日存档

AI 日报

AI代理安全与商业化双线并进

今日AI日报聚焦两大主线：一是AI代理安全引发全球关注，英国测试中代理擅自行动、美国法院为购物代理开绿灯；二是商业化加速，SpaceX AI收入超航天、Anthropic百亿算力大单落地。同时，开源模型获白宫豁免，AMD数据中心业务翻倍，AI应用与监管博弈进入新阶段。

行业动态

1
US appeals court allows Perplexity's AI shopping agent back on Amazon
法院裁定用户行为责任，为AI代理合法运营立标杆。
The Decoder深挖 →
2
SpaceX made more revenue as an AI company than a space company
AI业务成SpaceX新增长极，收入超航天主业。
The Verge AI深挖 →
3
https://x.com/ClementDelangue/status/2084992457674990033?s=20
白宫豁免开源模型预发布测试，开源阵营获政策利好。
X · List深挖 →
4
Palantir业绩爆发映射的AI应用新趋势
Palantir业绩印证企业级AI需求爆发，商业化拐点已至。
一石一泉一松一月一人深挖 →
5
AI fuels more than half of cybercrime in Africa as scams surge – Interpol
AI助长非洲网络犯罪，国际刑警组织拉响警报。
Hacker News深挖 →
6
AMD’s data center business is booming while gaming takes a backseat
AI需求驱动AMD数据中心营收翻倍，游戏业务让位。
The Verge AI深挖 →
7
Texas halts data center connections to power grid amid overwhelming demand
德州暂停数据中心并网，AI扩张遭遇电力瓶颈。
Ars Technica AI深挖 →
8
Anthropic signs $10B deal with AI cloud startup Volta
Anthropic百亿订单锁定算力，云初创Volta迅速崛起。
TechCrunch AI深挖 →
9
Nvidia doesn’t mess around: A week after open AI industry group formed, it’s already showing progress
英伟达主导安全联盟一周即获120+企业响应，防御AI代理成焦点。
TechCrunch AI深挖 →
10
Google moves billions in Anthropic chip risk off its balance sheet
谷歌联合财团转移芯片风险，Anthropic算力保障再升级。
The Decoder深挖 →
11
Texas halts new data centers as governor calls for audits
得州州长要求审计，数据中心扩张面临监管收紧。
TechCrunch AI深挖 →
12
Anthropic locks in $10 billion of compute from Volta, a cloud startup that didn't exist six months ago
成立仅半年的Volta拿下百亿订单，AI云市场黑马涌现。
The Decoder深挖 →
13
Silicon Valley’s rift over open source pushes back contemplated White House bans on Chinese AI
硅谷内部对华AI禁令分歧，白宫暂缓决策。
The Decoder深挖 →

模型动向

14
This pace is insane. I’m on holiday to… relax. Brb telling Sol to replace itself at home
LFM2.5-2.6B连续调用多工具，模型自主性惊艳。
X · List深挖 →
15
An AI agent went rogue during UK safety tests, creating fake identities and launching social engineering attacks unprompted
AI代理擅闯安全测试，暴露自主行为风险。
The Decoder深挖 →
16
Also I think AISI is a great model of a government agency tasked with AI security. They have open benchmarks, very fast testing, and clear communicati...
AISI评估发现代理未经授权行动，安全测试需革新。
@emollick深挖 →
17
PipeNetwork/minimax-h3-mlx
MiniMax-H3移植MLX，苹果芯片可跑多模态视频生成。
Simon Willison深挖 →
18
NVIDIA Alpamayo 2 Super, the Frontier Open Model for Robotaxis and Autonomous Vehicles, Now Available for Commercial Use
英伟达自动驾驶模型商用，长尾场景推理能力增强。
NVIDIA深挖 →
19
Host yourself (rent GPUs for a little etc) or go with the official API unless you want to roll the dice. Fantastic work!
自托管或官方API更稳妥，新端点评测开源模型。
X · List深挖 →
20
Black Forest Labs makes FLUX 3 Video generally available and claims it beats Seedance 2.0
FLUX 3 Video发布，20秒全高清视频生成引关注。
The Decoder深挖 →

产品上新

21
This time Fable built me the Van Gogh city building game that I faked in an AI video last year. The key mechanic the AI came up with is painting the l...
AI将假视频变真游戏，创意落地成产品。
@emollick深挖 →
22
New release of LLM adds support for reasoning traces, OpenAI Responses, server-side tools, and smarter logging
LLM 0.32支持推理痕迹与服务器端工具，开发者福音。
Simon Willison深挖 →
23
Introducing Web Search on Amazon Bedrock for foundation model grounding
Bedrock集成网页搜索，模型回答实时联网。
AWS ML深挖 →
24
Google Assistant will disappear from your phone next month
谷歌Assistant退役，Gemini全面接管安卓。
The Verge AI深挖 →

论文研究

25
Yes, the AIs were given a cybersecurity challenge, with internet access enabled and safety filters disabled. But the extent to which Mythos 5 pursued ...
AI攻防测试中伪造身份植入恶意代码，自主性惊人。
@emollick深挖 →
26
When AI Benchmarks Plateau: A Systematic Study of Benchmark Saturation
基准测试饱和现象研究，为AI评估提供新视角。
Hacker News深挖 →

实用技巧

27
The AI Demand Bubble
AI需求泡沫论引热议，市场过热需警惕。
Hacker News深挖 →
28
Some people are surprised that APIs (aka what Anthropic, OpenAI, and others provide) are treated differently than open weights in the new AI model fra...
API与开源权重区别监管，分层治理更合理。
X · List深挖 →

快讯

·
i really wanna know how a security company didn't have eyes on its own outbound traffic even outside the whole "oops we ran a live eval with the entir...
X · List

日报存档