AI News 2026-09-02 — Top 10 AI Stories

Curated AI news digest for 2026-09-02 (timezone Asia/Shanghai).

  1. How AI-native companies turn workflows into operating capability — OpenAI News. Basis, Clay, and Exa Labs use AI agents to improve onboarding, account management, and developer integrations. See what enterprise leaders can apply.
  2. Introducing agentic video understanding with Gemini — Google DeepMind.
  3. NVIDIA and CrowdStrike Strengthen Agentic Cybersecurity Frontier — NVIDIA Generative AI. “We’re at an inflection point in cybersecurity,” Jensen Huang told a sold-out crowd at CrowdStrike’s Fal.Con 2026 in Las Vegas Tuesday. Attacks are now automated. Defense has to be, too. The NVIDIA founder and CEO joined CrowdStrike CEO and founder George Kurtz to announce Crowd…
  4. OpenAI Is About to Release Its First AI Model With ‘Critical’ Cyber Abilities — WIRED AI. The company will give select partners early access to its Astra AI model—so they have time to shore up their defenses.
  5. Anthropic launches Claude Fable 5.1 and says it’s up to 45 percent cheaper for agentic work — The Verge AI. Anthropic says its newest AI models, Fable 5.1 and Mythos 5.1, address criticisms from customers about price, data retention, and overzealous safeguards. The company claims Claude Fable 5.1 offers stronger performance than Fable 5, but costs around 25 percent less typically and…
  6. OpenAI’s Astra model is on the way — and very good at breaking into computer systems — TechCrunch AI. OpenAI previewed the precautions it is taking as it prepares to release Astra, its newest, cyber-critical LLM.
  7. Claude最强Fable 5.1发布!8项屠榜,最高降价45%,反蒸馏机制上线 — 量子位. 无论你此前让Claude处理什么任务,Fable 5.1都能做更多,并且把最难的部分完成得更出色
  8. DeepSeek V4 多模态开源,我们把它的视觉链路拆了一遍 — 雷锋网. 图片不只被编码,还直接参与 Attention、MoE 与 Agent 推理。 作者丨郑佳美 编辑丨岑 峰 这不是一个全新的模型突然出现。早在 8 月 21 日,DeepSeek 就已经把它接入 API,当时外界只能看到结果:V4 开始能处理图片,而且在 ApexBench、Agents' Last Exam、Chartography 等多模态 Agent 基准上整体提升。 现在不一样了。 随着权重和参考推理代码公开,V4 的视觉部分第一次可以直接拆开看。Vision Encoder 怎么处理图片,Aligner 怎么把视觉特征压进语言空间,视觉 To…
  9. 万字长文拆解DeepSeek V4 Pro与Harness:从后训练到「代理自进化」,更大的变化在开源框架里 — 雷锋网. 不管哪个时代,人都会遇到同一道题:无论宏大叙事还是人生决策,面前问题太复杂,资源和时间却不够,必须决定先做哪个、放弃哪个。(雷峰网) 从 1812 年黑格尔的辩证法,到马克思、列宁,再到 1915 年毛泽东在延安写下的《矛盾论》,一百多年里,这套思想最终落到了一个很实际的问题上:复杂事物中矛盾很多,但真正重要的是找到决定其他矛盾的主要矛盾。 今天的大模型行业也不例外,当前大模型行业面临的主要问题有以下这些: 1.模型能力的增长速度,同把这种能力转化成可靠产出的能力之间的矛盾(模型强,但产出物没价值)。 2.算力资源受限,同能力需求无限之间的矛盾(人们总…
  10. 我们用 Qwen 3.8-Max 做了一个 AI 大模型宇宙:效果竟然胜过 GPT5.6? — 雷锋网. Max长程Agent能力的背后,是精雕细琢的典范,还是算力黑洞的失控? 作者丨李娜 编辑丨岑峰 我让Qwen 3.8-Max从零搭建了一个3D大模型演进宇宙网站,又和GPT-5.6做了迭代对比。 (Qwen3.8-Max运行成果截图) 效果上限方面,Qwen 3.8-Max 展现了远超 GPT-5.6 的交付细节; 效率和成本代价方面,Max高完成度背后是低效,而且因为无法自主判断任务终点导致了额度耗尽中断; 适用场景方面, 追求快速迭代、低成本原型,GPT 依然是首选,但在不计成本追求极致性能的场景,Qwen 3.8-Max 提供了一个昂贵但有效的国…

Back to latest AI News