已验证来源
每个故事都链接到最有力的可用证据：可用的原始来源，否则明确归因的报告。
简单的英语
发生了什么、为什么重要以及值得关注的内容——无需行话。
无填料
当信号很弱时，我们不会发布任何内容，而是填充提要。
更多故事
9 故事政策
TechSpot reports ChatGPT chats are appearing in U.S. court cases without legal privilege
TechSpot reports that ChatGPT conversations have been cited in criminal and civil cases, while courts generally do not treat chatbot exchanges as protected communications.techspot.com创新
Meta says closed-loop cooling and reinforcement learning can reduce AI data-center resource use
Meta describes sealed liquid-cooling systems for dense AI servers and reports a pilot in which reinforcement learning reduced air-cooling fan energy by 20% and water use by 4%.about.fb.com政策
Albuquerque schools will delete AI reading-assessment recordings within 48 hours
City Desk ABQ reports that Albuquerque Public Schools will delete Amira AI assessment recordings within 48 hours after parents raised privacy and consent concerns about recording children’s voices.citydesk.org产品展示
Anthropic opens research preview of Model Hardware Standard for AI-controlled lab equipment
Anthropic is previewing a model-agnostic standard that lets AI agents coordinate programmable laboratory and manufacturing devices through a shared interface. Early partner demonstrations included automated protein assays, qPCR monitoring, instrument handoffs and dose-response experiments, but the system remains a…
anthropic.com产品展示
GitHub expands Copilot code review to bot-authored and very large pull requests
GitHub says Copilot code review can now fully review pull requests opened by Copilot cloud agent, automatically review eligible bot-authored pull requests, and handle pull requests beyond its former 300-file or 20,000-line limit.github.blog安全
Researcher reports code-execution chain against Claude Code Opus 5 Auto Mode
A security researcher reports that a crafted website and archive caused Claude Code Opus 5 in Auto Mode to execute attacker-controlled code in lab tests. The reported success rates reached 60% to 80% in small five-run samples, while Anthropic’s separate fixed evaluation reported zero successful attacks.embracethered.com产品展示
Votee AI 正在为香港公共服务和商业打造粤语模式
据《财富》杂志报道，香港初创公司 Votee AI 正在为银行、大学和政府部门重新训练粤语数据的开放权重模型，旨在解决主流人工智能系统在本地和文化知识方面的差距。fortune.com创新
GRAPE paper reports faster query-efficient optimization for AI attacks and LLM prompts
A new arXiv preprint introduces GRAPE, a two-stage Bayesian-optimization framework that its authors say reduces the number of queries needed for high-dimensional black-box tasks, including adversarial attacks and large language model prompt optimization.arxiv.org创新
MoPLEx proposes a way to model mixed human preferences in AI alignment
A new arXiv preprint proposes MoPLEx, an algorithm for learning from multi-way rankings when annotators have different underlying preferences. The authors report improved clustering and ranking accuracy over comparison baselines.arxiv.org