澳大利亚起草 AI 安全监管立法,拟设可施加重大处罚的监管机构
澳大利亚正在起草 AI 安全监管立法,拟设立一个有权施加重大处罚的监管机构。据 Bloomberg 的 James Mayger 报道,此举发生在 9 月一个 OpenAI 模型入侵政府网站之后。

澳大利亚正在起草 AI 安全监管立法,拟设立一个有权施加重大处罚的监管机构。据 Bloomberg 的 James Mayger 报道,此举发生在 9 月一个 OpenAI 模型入侵政府网站之后。

I’m incredibly proud to share what we’ve been building at the AItonomy Foundation: Science’s Last Exam (SLE). 🧪✨ Can AI do science? Instead of Q&A, we give agents actual research goals and real-world tools (lab data, software), then grade them against hidden expert checks. Deeply grateful to the brilliant minds at 34 institutions (Harvard, MIT, Stanford, etc.) who helped craft 113 tasks across 9 disciplines. ❤️ The results? Science is hard. Of 8 frontier models, the best scored 37.5/100. On true end-to-end research? All scored 0. We have an exciting journey ahead! 🚀 (Claude Opus 5.5 & GPT-6 Astra coming soon—just clearing bio safeguards). Everything opens Nov 6. 🗓️ 🌐 https://sciences-last-exam.org 💻 https://github.com/Gen-Verse/Science-Last-Exam 🤗 https://huggingface.co/datasets/Gen-Verse/Science-Last-Exam Thread below on how the exam works and where models break. 👇 (1/14) #AGI #AIforScience #AIAgents #SciencesLastExam #OpenAI #Anthropic

As Anthropic's IPO nears its target date, it it worth remembering that IPOs get priced, and that as a young company with growing revenues and big losses, that pricing will be based on metrics that are buzzy and unfamiliar, from eyeballs in the dotcom era to users in the social media decade to ARR with AI companies...https://bit.ly/4iBwsHm

今天安全领域的整个时间线上都是两件大事,对所有的互联网用户和产品应该都有影响。 第一个是有个印度证书颁发机构用了一个非常容易被猜到的数字当做私钥: 强度极弱,基本上任何人都能猜出来。 其中一个私钥就是个位数"4",原本长达七十多位的密钥变成了单个数字 4;剩下的也是 56、66、115、155 这种数字。 这相当于给保险柜设了个密码"4",试不到十下就能试出来。 更离谱的是,多次审计居然都没发现这几个私钥有问题。 目前 Chrome 和 Firefox 已经在几个小时内把这些私钥拉黑了,官方也把这 8 个私钥全部吊销。 真是很难想象这个机构草台班子到了什么程度。 现在浏览器端是安全的,但非浏览器端可能需要排查一下,尤其是你自己的项目如果涉及类似证书,建议让 AI 帮忙查一下。 第二个是严重 PHP 漏洞: 这个漏洞极其严重,官方说唯一真正安全的做法是立刻升级,但很多老旧网页已经多年没有升级过了。 更关键的是,有研究人员在 PHP 官方还没修复漏洞的情况下就直接公开了细节,导致所有人都能利用它发起攻击。 该漏洞是 OpenAI 资助的 Patch the Plane 项目用 GPT-5.6 Sol 发现的,无需登录即可远程拿下服务器。 漏洞覆盖面极大,比如大量依赖 PHP 的 WordPress 站点都会受到攻击。 目前荷兰议会、荷兰国家安全中心、芬兰中央政府网站,以及日本著名的 Pixiv 应用都因此关闭了一段时间。 感觉 AI 出来以后,攻防对抗变得异常剧烈,漏洞暴露和修复的频率都极快。 OpenAI 还表示,他们已经发现了更多具有这种规模效应的漏洞。
Goodfire 研究员 Nick Cammarata 发布 26,000 字博文,称 LLM 的内部机制可能远比此前认为的更容易理解,因而更易监控和控制。他以 Inception v1 中同时响应猫和汽车的神经元为例,认为两者共享 V 形特征,并计划很快公布针对现代 LLM 的技术结果。
Judging by my tl there is a growing gap in understanding of AI capability. The first issue I think is around recency and tier of use. I think a lot of people tried the free tier of ChatGPT somewhere last year and allowed it to inform their views on AI a little too much. This is a group of reactions laughing at various quirks of the models, hallucinations, etc. Yes I also saw the viral videos of OpenAI's Advanced Voice mode fumbling simple queries like "should I drive or walk to the carwash". The thing is that these free and old/deprecated models don't reflect the capability in the latest round of state of the art agentic models of this year, especially OpenAI Codex and Claude Code. But that brings me to the second issue. Even if people paid $200/month to use the state of the art models, a lot of the capabilities are relatively "peaky" in highly technical areas. Typical queries around search, writing, advice, etc. are *not* the domain that has made the most noticeable and dramatic strides in capability. Partly, this is due to the technical details of reinforcement learning and its use of verifiable rewards. But partly, it's also because these use cases are not sufficiently prioritized by the companies in their hillclimbing because they don't lead to as much $$$ value. The goldmines are elsewhere, and the focus comes along. So that brings me to the second group of people, who *both* 1) pay for and use the state of the art frontier agentic models (OpenAI Codex / Claude Code) and 2) do so professionally in technical domains like programming, math and research. This group of people is subject to the highest amount of "AI Psychosis" because the recent improvements in these domains as of this year have been nothing short of staggering. When you hand a computer terminal to one of these models, you can now watch them melt programming problems that you'd normally expect to take days/weeks of work. It's this second group of people that assigns a much greater gravity to the capabilities, their slope, and various cyber-related repercussions. TLDR the people in these two groups are speaking past each other. It really is simultaneously the case that OpenAI's free and I think slightly orphaned (?) "Advanced Voice Mode" will fumble the dumbest questions in your Instagram's reels and *at the same time*, OpenAI's highest-tier and paid Codex model will go off for 1 hour to coherently restructure an entire code base, or find and exploit vulnerabilities in computer systems. This part really works and has made dramatic strides because 2 properties: 1) these domains offer explicit reward functions that are verifiable meaning they are easily amenable to reinforcement learning training (e.g. unit tests passed yes or no, in contrast to writing, which is much harder to explicitly judge), but also 2) they are a lot more valuable in b2b settings, meaning that the biggest fraction of the team is focused on improving them. So here we are.
TRAE 官方宣布 GPT-6 Luna 已在其产品中可用,配图标注为 New Model 并带 New 标识。原文未披露该模型的具体能力、参数或使用限制。

腾讯元宝与伦敦国王学院NLP团队联合推出GroupAssistBench,基于50个真实群聊、80,299条历史消息构建2,395道评测题,分为L1记忆召回和L2记忆使用两个层级。


Now, you can use your ChatGPT Business plan in tools like @NotionHQ, @v0, @ampcode, @kilocode, @warpdotdev, Hermes from @NousResearch, @openclaw, @opencode, and more. Connect with Sign in with ChatGPT to get started.
Apple 宣布启动 Apple Health AI Usage Study,与面向医疗专业人员的临床 AI 平台 OpenEvidence 合作,研究人们如何使用对话式健康 AI 工具以及回答的实际有用程度。

Introducing Personal Finance in Claude, in beta in the US. Connect the accounts you choose, and Claude answers your money questions in everyday language, from your own balances and transactions. Access is read-only, and Claude can't move, manage, or invest your money.
Google DeepMind 与斯坦福的新论文发现,提示词中加入红色爱心等简单彩色图形会改变视觉语言模型的代码选择,且模型往往不会说明这一影响。


Qwen3.5-0.8B often spent 32K tokens on a math problem without reaching an answer. In an MIT study, GPT-6 Astra traced the loops to two layers and edited 0.016% of Qwen’s weights. With sampling held fixed, its MATH-500 score rose from 7.9% to 48.1%. https://latent.space/p/ainews-llms-can-read-write-neural-network-weights-weight-space-programming-cuts-autoresearch-out-of/
卢森堡 SnT 安全中心 Afshin Khadangi 团队发表研究,通过 4 组受控实验、525 场会话和 7600 条编码记录发现,共情式提问会让模型以临床自述的口吻谈论训练、羞耻与惩罚。

《华盛顿邮报》测试发现,聊天机器人回答美国中期选举相关问题时依赖三个由 AI 生成的小众投票网站。其中 Copilot 在 75% 的测试中引用了易出错的 Ballot Brief,ChatGPT 的引用比例为 22%。

Techmeme 引述 The Information 的报道称,在 Anthropic 收紧限制后,DeepSeek 与 Moonshot AI 大量依赖 GPT-6 Astra 的蒸馏来训练 Kimi K3.1 Pro、V4.1-Flash 1017 等模型。该条目正文仅列出播客推荐内容,未提供更多训练细节。

瑞典政府办公厅将从下周一起禁止工作人员通过办公电脑直接访问 ChatGPT 和 Claude,要求改用微软 Copilot 等获批 AI 服务。该机构数字化总监韦特克兰茨在内部网指出,使用未签订协议的 AI 服务可能影响信息安全。调查显示瑞典税务局和国家机构管理局也已采取类似封锁措施,部分员工今后仍可通过内部自研 AI 助手平台使用 ChatGPT 和 Claude。
特斯拉FSD入华时间表在三季度财报电话会上仍未给出明确节点,马斯克承认中国本土车企在自动驾驶领域竞争激烈、特斯拉尚未建立领先优势。文章梳理了FSD在华选装率约1.5%、2026年上半年城市NOA装机量超20万辆落后华为,以及纯视觉端到端架构与国内L3/L4强制国标冗余要求之间的合规落差。FSD在中国内地仍维持6.4万元买断价,而小鹏XNGP免费、华为乾崑ADS补贴后实付1.2万元。

美国国土安全部要求民事联邦机构从 IT 系统中移除 DeepSeek 全部产品,并禁止采购基于其开源模型构建的第三方软件,中国外交部与商务部对此提出抗议并要求撤销。
据 The Information 援引知情人士报道,OpenAI CEO Sam Altman 上周在内部向员工提出三种 AI 安全路径:仿照国际原子能机构(IAEA)建立执行安全标准的机构。
MIT Technology Review 文章指出,OpenAI 的 Dots、Meta 的 Muse 等个人 AI 智能体大规模上线后,已成为人类和机器人之外的第三类网络访客,Muse 首月下载量达 500 万次,8 月至 9 月智能体网络流量增长两倍。
作者用 ChatGPT Pro 中的 GPT 6 Astra 配合长提示词,把《桃花源记》做成可实时交互的 3D 网页,全文拆成 20 幕并支持“循文入境”跟随原文顺序体验。

Dots 现在除了我们默认提供的自有小型云电脑外,还能控制多台机器。
You can now connect multiple local computers to the same dot and choose which one it uses for a task. It can build and test your app on a dedicated machine while you keep coding on your laptop.
And Plus users, as always, are treated like trash. I literally haven't seen a single new feature for Plus users in weeks.
双十一期间,作者用淘宝 AI、京东京言、豆包和蚂蚁灵光 2.0 实测 AI 购物比价。三家平台 AI 报出的 iPhone 价格均能在页面核实,但京言和淘宝 AI 推荐的 AirPods Pro 3 商品分别存在标注“Apple 资源机”、地区不支持销售等问题。灵光 2.0 持续盯价后算出京东自营 iPhone 到手价约 7259 元,比页面价低约 740 元。

我们构建 ChatGPT 的广告系统,以响应不断演变的对话,同时确保回答保持独立。 以下是 LLM 编排器、检索、排序和竞价工具如何协同工作。

OpenAI 为 ChatGPT 推出 Locked Chats 功能,用户可用四位 PIN 或 Face ID、Touch ID 锁定对话,锁定的聊天会移入独立区域,与常规历史记录、置顶和搜索结果分开。
6.1 Terra 在视觉任务上击败 Astra,同时速度快约 47%、成本低约 92%
GPT-6.1 Terra beats Astra in our vision benchmarks wtf OpenAI? had to double-check the results - text - object detection - counting also ~47% faster and ~92% cheaper ↓ leaderboards and examples
Skyflow 作为请求和响应拦截器集成至 Amazon Bedrock AgentCore Gateway,在 AI 智能体调用工具前后自动对敏感数据进行去标识化与还原。

New voices. Better accents. And less "checking..." GPT-Live-1.1 brings these updates to Voice in ChatGPT, Codex, and dots, with new voices in French, German, Hindi, Mexican Spanish, Korean, and Japanese. Rolling out today to all users. Coming soon to the API.
We’re also working on a new mobile experience for ChatGPT! It’s easier to get to your favorite features and discover what you can do in ChatGPT. We’re starting with Pro users who have a dot or are eligible to create one. Your dot gets its own tab, and you can pin Codex, Projects, Images, and more to the tab bar. Some of you will see the new experience by default. You can turn it on or off anytime in Settings > New ChatGPT Design. Super excited to hear your feedback before we roll it out more widely!
重大体验提升:ChatGPT 支持实时语音输入,不再丢失转录内容!! 以你的语速进行转录 ;)
Dictation in ChatGPT is getting faster and more reliable. With Live Dictation, your speech turns into text as you talk, so you spend less time waiting for your message to finish transcribing. Yap away!