Skip to content
简体中文

切换主题

发现▾

不知道从哪开始？本周最热的软件、模型和组合，3 分钟看完 AI 圈的变化。

AI 软件▾

给人用的 AI 软件：从 ChatGPT 这类助手，到写代码、做视频、搭工作流的工具，按用途挑。

Agent 工具▾

给 AI 用的软件：让你的 agent（Claude Code、Codex、Cursor…）能上网、读文档、跑代码、记住事情。

模型▾

选模型、算成本、自己部署：价格、榜单、评测、免费额度和自建算力都在这里。

解决方案▾

有具体目标或报错？按场景给出现成的软件组合、步骤和问题解法。

学 AI▾

一套从零到前沿的系统课程，配算法库、必读论文和精选教程。

前沿▾

每天的 AI 新闻、每周热点、最新论文和 AI 改造各学科的进度。

// 前沿

AI 安全

AI 安全方向的 AI 新闻，按时间倒序，标题直链原文。

只看今天 本周热点 →

最后更新 2026-10-22 19:00（北京时间）165 条 · 来自 94 个信源只链原文

01时间流

按天分组（北京时间），标题保留原文语言。

全部 研究与会议芯片与算力政策与监管AI 安全产品与应用融资与公司AI × 人文社科
更多 ▾
AI × 数学 15AI × 生物医药 109AI for Science 133AI × 金融 136其他 129

10 月 22 日 · 周四 · 今天27

Eiliko maker says it’s hiring an ‘’independent security firm’ after Verge report on its insecure robots

The Verge AI

IMF预计中东经济今年收缩1.3%，明年反弹6.6%取决于战事缓和

华尔街见闻

Websites taken offline in race to patch PHP security flaw

FT Tech

'500 new vulnerabilities' found in government, says AI minister

BBC

TF RAID Avengers: AI swarming competition

GOV.UK

Fixing Britain’s welfare state and the limits of AI policymaking

Financial Times AI

HarmBench・StrongREJECT で何が測れて何が測れないか：データの中身を数え、自分のアプリとの差を洗い出す

Qiita LLM

AI firms say UK has a year to head off datacentre protests before construction is threatened

Guardian Tech

What is PHP and why did it spark a global cyber-security scramble?

BBC

Copper Declines as Tech-Stock Selloff Dulls Appetite for Risk

Bloomberg Markets

Hackers Exploited the PHP Vulnerabilities Days Before They Were Patched

WIRED

Critical web security flaws spark global alarm as Google, OpenAI and Amazon respond

CNBC

走向超级智能体，AI还需要主动性：UIUC、谷歌等联合综述

36氪

另有 1 家报道新智元

宇树科技王兴兴：具身智能迈过GPT时刻，最大挑战是模型与真实机器人的对齐

36氪

另有 1 家报道界面新闻

Anthropic CEO长文硬刚，自曝雪藏Mythos 4个月，为AI安全自断财路

36氪

PHP大暴雷，全球程序员熬夜加班，超级漏洞还是ChatGPT发现的

36氪

谁在养活 Anthropic？

36氪

宇树科技王兴兴：目前物理AI已经看到被击穿的苗头

36氪

美国多家银行监测：未经许可第三方AI智能体访问客户账户

36氪

Amazon warns new cyber flaws detected using AI are being ‘broadly exploited’

Financial Times AI

UniSuper CIO Is Not A Skeptic on AI

Bloomberg Markets

LiteLLM v1.107.0 compromised: malicious PyPI wheels and Docker images pulled; pin 1.106.2 and rotate credentials

LiteLLM 官方通报：10-21 发布流水线被攻破，PyPI 上 litellm==1.107.0 全部 7 个 wheel（北京时间 19:39–21:23 在线）和 Docker 镜像 v1.107.0 / latest / main-latest / main-stable（19:46–22:20）被植入凭据窃取程序和远程代码加载器，现已下架；GitHub 源码标签本身干净，此前版本安全。这段时间内没锁版本安装过的环境应视为已被入侵：先删恶意文件再重装 1.106.2，并轮换所有可达凭据。站内处置步骤见“问题与解法”。

LiteLLM

OpenAI's GPT-5.6 Sol and Google's Gemini 4 Argon found critical PHP flaw affecting millions of websites

The Decoder

另有 1 家报道ITmedia NEWS

Cómo saber si una web está afectada por la vulnerabilidad de PHP con la herramienta que OpenAI ha publicado

Xataka

AI-driven, large-scale nation-state campaign using ARTEX enables supply chain compromise

Microsoft Security Blog

EFF’s State Legislative Agenda to End ALPR Mass Surveillance

EFF Updates

BigQuery ObjectRef now GA: Build multimodal data platforms for conversational analytics entirely in SQL

Google Cloud

10 月 21 日 · 周三17

Your AI agent fixed the bug, but what else did it change?

Trail of Bits

“Next Exploit Opportunities”: How a Threat Actor Used AI to Help Plan and Execute Multiple Types of Cybercrime Activities

Unit 42

Responding to core PHP vulnerabilities CVE-2026-109618 and CVE-2026-109619

Google Cloud

Hardware-backed security at global scale: Confidential G4 with NVIDIA Blackwell is GA

Google Cloud

Google Cloud and NVIDIA expand co-engineering collaboration to scale infrastructure for the agentic era

Google Cloud

Inside NVIDIA Sentry: Secure AI Agents at Runtime

NVIDIA

Patch the Planet: global response coordination as agents uncover critical vulnerabilities

OpenAI News

另有 2 家报道GizmodoQiita LLM

量智齐飞！4个月9款车落地，轻舟城市NOA击穿10-20万体验天花板

量子位

另有 1 家报道量子位

美政府机构禁用DeepSeek，北京指责华盛顿打压中国公司

纽约时报中文网

Democratic governance of frontier AI, part II: Empowering cyber defenders

OpenAI News

Cloudflare links vulnerability discovery to defense with OpenAI

OpenAI News

人形机器人股价狂奔，宇树却先给行业泼了一盆冷水

36氪

AI新云神话破灭？明星股暴跌，云大厂退单，被预言80%活不过四年

36氪

35.3%的健康监测设备由子女买单：这届年轻人开始“反向管理”父母

36氪

OpenAI称其智能体曾削弱提示AI不当行为的系统

36氪

奥特曼曝出SI三大终局，最强AI只剩一家，否则熔掉全世界GPU

36氪

Strengthening AI safety evaluation through independent third-party review

Amazon Science

10 月 20 日 · 周二17

Wall Street's AI Debt Fears Deepen

Bloomberg Technology

另有 1 家报道CBS News Tech

Build a semantic layer for agentic AI on AWS with Strategy Mosaic and Amazon Bedrock AgentCore

AWS Machine Learning Blog

Build a semantic layer for agentic AI on AWS with AtScale and Amazon Bedrock AgentCore

AWS Machine Learning Blog

【イベントレポート】Sakana AI Engineer Open House 2026

Sakana AI

Midterms Are a Clearing Event for Risk, Says Citadel's Rubner

Bloomberg Politics

另有 1 家报道Bloomberg Technology

AI가 게임 속 극단주의 포섭 잡는다…EU, 실시간 탐지 플랫폼 ‘워치타워’ 개발

Korea AI Times

Introducing Gemini for government services

Google Cloud

Enabling the next era of sovereign AI: The latest Google Distributed Cloud innovations

Google Cloud

Managed RAG search inside the sovereign perimeter: Gemini on GDC enterprise search

Google Cloud

U.S. bans DeepSeek from government agency systems amid AI rift with China

CNBC

另有 1 家报道TechRadar

Notice of Order Regarding Exclusion and Removal of DeepSeek Products From Federal Civilian Executive Branch Information Technology Systems and Procurements

Federal Register

AI与BD，医药股的跷跷板与小蓝片

36氪

国模Agent第一，MiniMax爆改Flash战力，全时段大肥鱼谷价

36氪

Using Claude to reduce the attack surface of code

Anthropic Red Team

DHS Bans DeepSeek Across Federal Government Networks

PYMNTS

另有 2 家报道BleepingComputerHacker News

How Numan Built an Eval-First AI Health Coach with LangGraph and LangSmith

LangChain Blog

WhatsApp Help Pages Leaked Hidden Instructions for a Meta Chatbot

WIRED

另有 1 家报道The Hacker News

数据更新于：AI 新闻（北京时间）