AI 每日快讯
AI 产品、模型、开源工具和官方动态的时间流。保留历史记录，按分类、日期和标签继续筛选。
原文摘要：OpenAI introduces the Australian Youth Safety Blueprint, a six-pillar roadmap for safer AI experiences that protect and empower young people. 来源：OpenAI 官方动态。建议继续查看原文，重点核对它影响的工具入口、成本、风险和真实使用场景。
原文摘要：Jina AI has released jina-ocr-v1, a visual document parser that converts PDFs, scans, tables, charts and invoices into Markdown. The model has 3.4B total parameters, with about 570 来源：MarkTechPost。建议继续查看原文，重点核对它影响的工具入口、成本、风险和真实使用场景。
原文摘要：You’re deploying a model on a system. It starts up, prompts are getting responses. Now the hard question: Is this fast? Your instincts might lead you to send... 来源：NVIDIA 开发者 动态。建议继续查看原文，重点核对它影响的工具入口、成本、风险和真实使用场景。
PrismML 发布 Ternary Bonsai 2 27B，这是 Qwen3.8 27B 的三值权重版本。根据原始信息，模型体积为 5.93 GB，而 FP16 版本为 53.80 GB，官方称在 20 项基准上保留了父模型平均表现的 98.2%。它支持文本和图像输入，上下文长度 262K，采用 Apache 2.0 许可，并演示了驱动 Cline 编码场景。这意味着在消费级显存或单卡环境下运行接近 27B 级别的模型变得更具可行性。值得关注的是，量化压缩长期面临能力损失，而 98.2% 的保留率如果可复现，会明显改变本地部署的性价比。影响的是本地推理玩家、边缘部署团队和成本敏感的编码 Agent 用户。下一步可以下载模型，在相同基准上复测，并重点验证长上下文和图像输入下的实际表现。
原文摘要：California Governor Gavin Newsom signed an executive order seeking independent auditors inside AI labs and a "kill switch" for AI models. An expert panel has two months to 来源：The Decoder。建议继续查看原文，重点核对它影响的工具入口、成本、风险和真实使用场景。
原文摘要：Three security researchers used Anthropic's Claude models to break into OpenAI's internal systems through its community forum in less than 72 hours. According to the team, 来源：The Decoder。建议继续查看原文，重点核对它影响的工具入口、成本、风险和真实使用场景。
原文摘要：Migrate a multi-model healthcare AI agent from self-managed Amazon ECS with AWS Fargate to Amazon Bedrock AgentCore runtime, preserving triple-model orchestration and vector-enhanc 来源：AWS Machine Learning 动态。建议继续查看原文，重点核对它影响的工具入口、成本、风险和真实使用场景。
原文摘要：Internal emails and sworn testimony undercut OpenAI and Microsoft's fair use defense. A Microsoft director described the practice as the "largest theft of labor in human h 来源：The Decoder。建议继续查看原文，重点核对它影响的工具入口、成本、风险和真实使用场景。
原文摘要：Deploy production-ready Hugging Face models on Amazon SageMaker AI using six open-source agent skills. Point a coding agent at a model and get back a real-time endpoint with the ri 来源：AWS Machine Learning 动态。建议继续查看原文，重点核对它影响的工具入口、成本、风险和真实使用场景。
原文摘要：For the first time, Anthropic is releasing metrics on how it builds its own AI. Claude already "leads" 26 percent of the work on future models, up from under one percent i 来源：The Decoder。建议继续查看原文，重点核对它影响的工具入口、成本、风险和真实使用场景。
原文摘要：AI models think out loud today, but Google Deepmind says that transparency is at risk. The article Visible chains of thought are a safety advantage for AI, but that transp 来源：The Decoder。建议继续查看原文，重点核对它影响的工具入口、成本、风险和真实使用场景。
原文摘要：DoorDash built a multi-agent LLM system to automate stale feature flag cleanup across more than 60,000 flags and 623 代码仓库. The 工作流 combines live experimentation data t 来源：InfoQ AI ML Data Engineering。建议继续查看原文，重点核对它影响的工具入口、成本、风险和真实使用场景。
原文摘要：OpenAI has introduced Astra for Law, a version of its GPT-6 Astra model built for legal work. The article OpenAI takes aim at the legal market with Astra for Law appeared 来源：The Decoder。建议继续查看原文，重点核对它影响的工具入口、成本、风险和真实使用场景。
原文摘要：Amazon SageMaker HyperPod Inference Gateway is a Kubernetes-native, GPU-aware routing add-on for Amazon EKS. It uses real-time GPU signals to send each inference request to the bes 来源：AWS Machine Learning 动态。建议继续查看原文，重点核对它影响的工具入口、成本、风险和真实使用场景。
原文摘要：WSO2 has announced the general availability of WSO2 Agent Manager, an open-source platform designed to provide centralized governance, identity management, security controls, and o 来源：InfoQ AI ML Data Engineering。建议继续查看原文，重点核对它影响的工具入口、成本、风险和真实使用场景。
原文摘要：42 Fellows of the Royal Society, including Fields Medal winners Martin Hairer and Peter Scholze, warn of existential AI risks in an open letter. Leading models have solved 来源：The Decoder。建议继续查看原文，重点核对它影响的工具入口、成本、风险和真实使用场景。
原文摘要：Which open-source harness works with Ollama, LM Studio, or llama.cpp? 11 verified picks with licenses and setup rules. The post Best Open-Source Agent Harnesses for Local LLMs in 2 来源：MarkTechPost。建议继续查看原文，重点核对它影响的工具入口、成本、风险和真实使用场景。
原文摘要：Experts from the US and China want to prevent AI systems from making autonomous decisions about deploying nuclear weapons. The article US and China experts push for shared 来源：The Decoder。建议继续查看原文，重点核对它影响的工具入口、成本、风险和真实使用场景。
原文摘要：The Instinct-to-Instinct model that is the talk of Silicon Valley presents too small a future for personal AI. With Fulcra’s multiplayer capabilities, people can get agent to agent 来源：AI 资讯。建议继续查看原文，重点核对它影响的工具入口、成本、风险和真实使用场景。
阿里 Qwen 发布 Qwen3.8-Omni-Flash，这是一个支持 100 万上下文的 omni-modal 模型，围绕 Agentic 音视频理解与工具使用设计，官方称在 OmniVideoBench 上减少约 45.7% 的 token 消耗。它值得关注的点有两个：一是把音频、视频理解和任务规划、工具调用放进同一个模型，二是长上下文叠加 token 效率优化，直接关系到 Agent 处理长视频或多模态任务的成本。影响的主要是做视频理解、会议分析、多模态 Agent 的开发者，以及需要处理长音视频素材的内容团队。验证方式：通过 Qwen 官方渠道申请或调用该模型，用一段带语音的长视频测试摘要与工具调用，记录实际 token 用量和延迟，再与现有方案做同任务对比。
原文摘要：Building an AI prototype is easy, but operating autonomous agents at scale requires production-grade tooling. Salesforce Agentforce bridges the gap from "vibe coding" to enterprise 来源：MarkTechPost。建议继续查看原文，重点核对它影响的工具入口、成本、风险和真实使用场景。
原文摘要：OpenAI has released a disclosure framework for model misalignment during its lifecycle. Employees can flag potential issues, prompting technical staff to label incidents. The initi 来源：InfoQ AI ML Data Engineering。建议继续查看原文，重点核对它影响的工具入口、成本、风险和真实使用场景。