OpenAI 按下暂停键：AI 安全关卡正从部署前移到训练
OpenAI 暂停前沿模型强化学习训练两周：当 Astra 逼近关键网络安全门槛，安全关卡从部署阶段前移到训练阶段。
OpenAI Pauses RL Training: Cyber-Safety Gates Move Earlier
OpenAI paused frontier RL training as Astra neared critical cyber capability, moving safety gates into the training phase.
AI Daily – 2026-08-19: DeepSeek Peak-Valley Pricing
DeepSeek introduces peak-valley API pricing, NVIDIA backs the OpenAI Ohio AI factory, and Unitree lists on the STAR Market.
AI 早报-2026年8月19日
每日精选全球 AI 资讯，快速掌握行业趋势 🔥 今日亮点（Top 3） 1. 年轻人首选B级闪充轿车秦MAX正式上市，售价10.99-14.39万元 核心洞察：AI 行业持续火热 来源：https://www.qbitai.com/2026/08/474907.html 2. 网易传媒发布”蜜蜂AI” ：从工具到伙伴，让AI更懂人 核心洞察：新产品密集推出，竞争加剧 来源：https://www.qbitai.com/2026/08/474857.html 3. π0引用的中国团队，又出手了：世界仿真器新作发布 核心洞察：新产品密集
AI Assistant Monetization Replays the Internet's Old Playbook
AI assistants start charging, replaying the internet's old playbook: subscriptions, ads, and commissions all hit walls.
Claude 隐形水印全量上线：AI 内容溯源进入「可证安全」时代
Anthropic 自 8 月 2 日起为 Claude 输出全量打上隐形水印、文件附带 C2PA 溯源签名。AI 内容溯源正从「经验安全」跨入「可证安全」.
Claude Invisible Watermark: AI Provenance Goes Provable
Anthropic now watermarks every Claude text output and signs files with C2PA metadata. Why AI provenance just turned from empirical to provable.
AI 批量生成时代：审美降级之下，品味成为新的护城河
AI 批量生成让内容泛滥成灾，护城河从「会生成」转向「品味」：品牌开始为审美品控单独付费，Midjourney 靠审美守住份额，IP 授权把风格工业化。
AI Slop Era: Why Aesthetics Is the New Competitive Moat
As AI slop floods, the moat moves from skill to taste: brands pay for aesthetic QC; Midjourney proves taste is a business.
美国致信35国「选边站」：全球AI供应链进入阵营化时代
美国国务院起草致35个伙伴国的信函，要求其在中美两大AI阵营中「选边站队」，明确「不能两边都要」。全球AI供应链阵营化正从信号走向落地，这是产业生态分叉的关键时刻。
US-China AI Blocs: 35 Nations Forced to Pick a Side
Washington drafted a letter pressing 35 partner nations to pick one AI bloc, ending dual US-China supply chain membership.