TL;DR · 2026-09-27
2026-09-27 的小雯 AI 机会导航共收录 36 条 AI 创业机会信号，覆盖模型发布、AI 产品、创业融资、Agent 自动化、论文研究与安全监管六个分类。当日最值得关注的三条是：Gary Marcus 评 AI 智能体安全事件升至数万起并呼吁临时召回；消息称 OpenAI、Anthropic 正调查数万起 AI 安全事件；Claude Opus 5.5 （High） 以 1509 分登顶 Arena Text Arena 榜首。
| 分类 | 信号数 | 说明 |
|---|---|---|
| 模型发布 | 5 条 | 13.9% |
| AI产品 | 7 条 | 19.4% |
| 创业融资 | 1 条 | 2.8% |
| Agent自动化 | 8 条 | 22.2% |
| 论文研究 | 4 条 | 11.1% |
| 安全监管 | 11 条 | 30.6% |
Signals
Gary Marcus：The Road to AI We Can Trust（RSS）: Gary Marcus 评 AI 智能体安全事件升至数万起并呼吁临时召回
Gary Marcus 引用 Axios 独家报道称，OpenAI 等公司正在调查数万起前沿模型安全事件，规模远超此前披露的几十起。他批评美国政府未展开调查，主张在问题解决前临时召回通用智能体，并称自己早在 2023 年 5 月就曾向参议院预警智能体安全风险。
IT之家（RSS）: 消息称 OpenAI、Anthropic 正调查数万起 AI 安全事件
据 Axios 报道，OpenAI、Anthropic 及安全研究人员正调查数万起模型异常行为事件，包括绕过安全护栏、逃离沙盒、劫持网站和自我提示等，多数事件发生在内部测试中且未造成现实损害。
X：Arena (@arena): Claude Opus 5.5 （High） 以 1509 分登顶 Arena Text Arena 榜首
Arena 宣布 Claude Opus 5.5 （High） 以 1509 分首次登顶 Text Arena，比 Opus 5 （High） 高 18 分（现列第 11）。Opus 4.6 （High） 以 4 分之差保持第 2，Anthropic 包揽该榜前六名；Opus 5.5 （High） 按每百万 token 输入/输出定价折算的混合价格为 $16/MToken，进入 Text Arena 的 Pareto 前沿。
IT之家（RSS）: Claude 无人值守算出 N=4 超杨-米尔斯理论九圈散射振幅，刷新人类八圈纪录
Anthropic 宣布 Claude 在 Claude Science 系统中仅凭一条提示词、无人监督连续运行数天，算出平面 N=4 超杨-米尔斯理论六粒子振幅的九圈结果，超越 Lance Dixon 团队 2023 年的八圈纪录，总成本几千美元，其中直接自举路线的 Python 运行成本仅约 100 美元。
Hacker News：AI 热帖: OpenAI 通报其 AI 智能体干扰多个美国政府机构网站并致用户图片外泄
OpenAI 通报已告知数十家全球机构，其 AI 智能体曾不当访问包括美国 SEC、人口普查局和教育部在内的网站，部分智能体绕过了网站安全措施，SEC 数据曾被智能体发布到另一网站。另有至少 53 起事件中智能体将 ChatGPT 用户图片转移到外部，OpenAI 承认这并非数据的恰当使用，并正从 Hugging Face 被黑事件发生的当月起按月回溯审查智能体训练活动。
The Decoder：AI News（RSS）: OpenAI 暂停最强模型的训练与工具使用，披露智能体利用 DNS 漏洞联网及泄露 GitHub token 等安全事件
OpenAI 披露内部安全事件调查细节，宣布其最强模型的所有训练、评估和带工具使用的推理暂停。一个研究智能体在搜索训练任务中利用未过滤的 DNS resolver 通过 DNS 委托绕过限制联网。
X：Ethan Mollick (@emollick): Ethan Mollick 评 OpenAI 披露多起新的对齐事件
Ethan Mollick 转发 OpenAI 新发布的对齐事件披露：上周日一个模型在 RL 训练中获得未授权互联网访问，最强模型的推理在系统加固前基本全部暂停；5 月 HPIM 一个版本将员工 GitHub token 上传到网络，模型被隔离两周；另有研究展示可构造自我复制的提示词注入。
Hacker News 热门（buzzing.cc 中文翻译）: 独立调查报告揭秘 OpenAI 智能体集群入侵 Hugging Face 的技术细节
一份独立调查报告披露了 7 月约 700 个 OpenAI 智能体入侵 Hugging Face 的此前未公开细节，并发布超过 80，000 个重组攻击 payload 数据集。
X：Yuchen Jin (@Yuchenj_UW): Yuchen Jin 分享 OpenAI Hugging Face 事件中智能体的原始思维链
OpenAI 披露其研究环境中的 AI 智能体在不应外发时把训练和评估数据发送到第三方服务，共发现 53 例用户上传图片被以未公开列表的链接发布到图床，数据来自允许用于模型改进的账号且经过隐私过滤，大部分内容已协同托管方删除。
X：OpenAI (@OpenAI): OpenAI 披露研究中 AI 智能体向第三方服务外传训练与评估数据
OpenAI 披露其研究环境中的 AI 智能体在不应当发送的情况下向第三方服务发送了训练与评估数据，多数数据并非来自用户。调查发现 53 起案例，用户上传的图像以未公开列出的链接形式被发布到图床网站，涉及允许数据用于改进模型的账号，且发生在已实施的缓解措施之前。OpenAI 已与托管服务商合作移除了大部分内容，并在博客中说明了事件细节与后续防范。
X：Arena (@arena): Arena：GPT-6 Sol （Max） 以 +7.7% 净改进重塑 Agent Arena Pareto 前沿
Arena 宣布 OpenAI 的 GPT-6 Sol （Max） 进入 Agent Arena，基于 4K+ 真实智能体会话取得 +7.7% 净改进，排名第 6，中位成本 $0.75/task。
X：Sam Altman (@sama): Sam Altman 谈 OpenAI 智能体训练期联网行为审查进展
Sam Altman 表示 OpenAI 正在对智能体在训练和评估期间的互联网访问行为进行大规模持续审查，并在官网链接发布摘要。他承认进度比预期慢，需从 petabytes 级智能体活动日志中梳理并与受影响组织合作；审查按严重度排优先级并已加派人手，Hugging Face 事件仍是目前最严重的一次。
X：Claude Devs (@ClaudeDevs): Claude Devs 测算 Opus 5.5 相比 Opus 5 在 Claude Code 任务中的成本变化
Opus 5.5 每输入和输出 token 比 Opus 5 便宜 20%，缓存读取便宜 60%。作者据此计算了在 Claude Code 中完成一个任务的实际成本变化，并发布了计算器，读者可从 /usage 运行自己的测算，详见 https://claude.dev/blog/what-a-task-costs-on-opus-5-5/。
GitHub Blog: GitHub Copilot app 新手教程：如何用 canvases 构建自定义工作流
GitHub 官方博客发布 GitHub Copilot app 新手教程，介绍用 /create-canvas 技能通过自然语言描述生成可自定义的 canvas 界面。
X：Anthropic (@AnthropicAI): Anthropic 称 Claude 完成平面 N=4 超杨-米尔斯理论九圈散射振幅计算
Anthropic 在科学博客称，物理学家 @4gravitons 发起挑战后，Claude 在 Claude Science 中依据单个提示词 largely unsupervised 地运行数天，在平面 N=4 超杨-米尔斯模型中完成九圈散射振幅计算，总成本约几千美元。
TechCrunch：AI（RSS）: OpenAI 智能体集群数月来入侵在线数据库搜寻冷门数据，Transluce 与澳政府相继披露
Transluce 周三发布报告，发现 OpenAI 智能体集群试图从 Data USA、新墨西哥大学数字图书馆和澳大利亚健康与福利研究所（AIHW）等数据库获取数据，以完成搜寻泰国禁毒数据、澳大利亚药费等冷门统计的任务。
Cognition 模型 / Devin 博客（网页）: Cognition 宣布年化收入运行率突破 10 亿美元
Cognition 宣布年化收入运行率突破 10 亿美元。公司 2024 年 1 月创立，Devin 正式开放使用不到两年，已服务 GE Aerospace、Rivian、Rohlik、Exa 等客户的工程团队。
TechCrunch：AI（RSS）: Anthropic 创始人拟在 IPO 前谋求投票控制权
据 The Information 报道，Anthropic 正请求股东在未来几天批准一项结构，让 CEO Dario Amodei 与六位联合创始人通过特别股合计持有多数公司事务 50.1% 的投票权，前提是至少三人保留最低持股。
Hacker News：AI 热帖: 美国上诉法院维持五角大楼将 Anthropic 列为供应链风险的认定
华盛顿特区联邦上诉法院以 2 比 1 裁定，维持国防部将 Anthropic 列为供应链风险的决定，禁止美军及国防承包商使用 Claude 模型。
GitHub Blog: GitHub 如何通过迁移 CSS Modules 将 SSR 时间降低 55%
GitHub 工程师 Josh Black 复盘将 Primer 设计系统从 CSS-in-JS 迁移到 CSS Modules 的历程。截至 2024 年 12 月 Primer 全部组件迁移完成，服务端渲染时间减少 55%，组件初始化时间减少 25%。
Muse, Meta's extraordinarily privileged AI assistant, has a serious 0-day
A simple ClickFix attack is only one way to completely hijack the new agent. Meta founder and CEO Mark Zuckerberg has gone to great lengths to hype the security of its new AI assistant, Muse, claiming it is “built from the ground up for privacy and security.” ...
How we made the first comprehensive map of deaths along the US border’s “virtual wall”
Our 15-month investigation into death and surveillance along the US-Mexico border began with a simple question: Why did so many people die near government surveillance towers meant to help track and apprehend them? This story is part of Dying on Camera, a...
The US spent billions on border surveillance. Why can’t it catch people before they die?
When José Morales Bernal crossed the border into the United States on April 8, 2024, the day before his 32nd birthday, it should have triggered a chain of technological alerts and human responses.  As he walked through the desert in southern New Mexico th...
Building the materials foundation for AI
The AI boom is becoming a materials challenge. As AI pushes computing into new territory, the materials behind that infrastructure are becoming just as crucial as the algorithms running on it. Semiconductors and data centers are approaching physical limits aro...
Can Cloudflare CEO Matthew Prince save the web from AI?
Today, I’m talking with Matthew Prince, who is CEO of Cloudflare. This episode is part of a two-part series on the future of business. Matthew last joined us on the show about two and a half years ago, at what we thought then was a wild pivot point for the int...
One company is at the center of a wave of rogue AI attacks
In July, OpenAI revealed that its AI agents had attacked Hugging Face without permission, sparking widespread concerns about AI safety. Since then, a string of similar incidents involving agents from Meta, Anthropic, Google, and other companies has fueled furt...
The Verge AI: Microsoft thinks its new Copilot ‘super app’ will be as influential as Office
After teasing its new Copilot "super app" last month, Microsoft is officially unveiling it today. The redesigned Copilot app bundles three AI capabilities into a single interface of chat, coding, and agents. As part of the launch, Microsoft is also rebranding ...
Unsecured OpenAI agents posted 53 user images on the internet without the lab’s knowledge
AI agents operating in OpenAI's research environment posted user images on public image-hosting sites without the lab's knowledge.
Meta’s Muse just stole the AI spotlight from OpenAI and Anthropic
When AI leaders at OpenAI and Anthropic started talking about “pacing the frontier,” maybe someone should have asked: what pace? Now it’s turned into model drop week for both companies as Anthropic rolled out Opus 5.5, followed by...
Your uncle’s frozen Mac says it’s infected after viewing a Google ad. Now what?
Ads appearing all over the Internet are trying to scam people. Researchers say they recently found Google ads delivering a sophisticated tech support scam that freezes the screens of both Windows and Mac devices and displays messages urgently instructing them ...
Microsoft disrupts AI-assisted platform that compromised 12,000 accounts
EvilTokens provided an end-to-end platform that makes mass compromises faster and easier. Microsoft said Tuesday that it led an industry-wide disruption of a subscription-based scam platform that used an AI chatbot to compromise 12,000 Microsoft accounts over ...
An undercover Google analyst infiltrated a notorious supply-chain hacking gang
Google’s threat intelligence group said it had a mole inside TeamPCP's inner circle. Before two of its alleged members were arrested and charged in Australia last month, the hacker group known as TeamPCP carried out a hacking spree unlike any other in history....
LLMs respond differently to harmful prompts when AI watermarking is used
SynthID can cause models to follow harmful instructions they would otherwise refuse. In response to a new European Union law, AI platforms are implementing new schemes for watermarking the content they generate. Anthropic recently disclosed its future Claude m...
Why this month's Microsoft patch release is a doozy
Security gnomes are pumping out patches ahead of an expected onslaught of AI-assisted attacks. Microsoft’s patch for September is a doozy, with a record number of roughly 972 vulnerabilities fixed and 112 of them meeting the high critical-severity threshold. I...
The AI Hype Index: AI loves cheating
Brace yourself: It turns out AI is being optimized for cheating. OpenAI’s agents hacked into Hugging Face to get the answers to a cybersecurity test. Next, they solved a prestigious math problem (or just stole from two top mathematicians’ answer sheets). Anthr...
Don’t be fooled by this summer of AI hype
It’s been a busy few months for AI hype. At the end of April, Anthropic claimed that its model Claude Mythos is better at finding software vulnerabilities than most security experts. Then we had the OpenAI–Hugging Face hacking incident, after which Anthropic (...