Quanta Magazine发布于 08/20 22:04
Are We Thinking Correctly About AI Intelligence?
（翻译）我们对 AI 智能的思考正确吗？
Computer scientist Melanie Mitchell discusses why artificial intelligence doesn’t “think” or “reason” like humans, and how we can create better methods for measuring machine cognition.
人人都是产品经理发布于 08/25 14:33
AI 私聊场景的审核错位：为什么”不生成违规内容”被偷换成了”用户违规”
当你把一份包含社区舆情、用户投诉、行业争议事件的竞品调研材料丢给 AI 辅助分析，弹出来的是"内容违规，请遵守社区规范"——这一刻，产品逻辑和场景认知发生了根本错位。本文从一次具体的产品经理工作场景出发，拆解 AI 平台将"公域社区惩戒逻辑"套用到"私域工作场景"的结构性问题，并指出：平台完全有能力在不牺牲合规安全的前提下，做到不影响用户体验。
InfoQ 中文发布于 08/24 17:03
Cloudflare WriteGuard 为 MCP 服务器提供了精细化的安全控制
Cloudflare 推出了 WriteGuard（目前处于私有测试阶段），旨在为 MCP（模型上下文协议）服务器提供精细化的安全控制。
Google DeepMind Blog发布于 07/17 23:00
Introducing Gemini 3.5 Flash Cyber
（翻译）推出 Gemini 3.5 Flash Cyber
Google introduces Gemini 3.5 Flash Cyber, a lightweight cybersecurity model to find and patch vulnerabilities.
Google DeepMind Blog发布于 07/16 17:30
Our approach to bioresilience
（翻译）我们的生物弹性方法
Google DeepMind and Isomorphic Labs are sharing our joint approach to bioresilience and AI models.
Solidot发布于 08/25 19:12
微软画图和照片应用生成的图像嵌入了看不见的水印
Solidot是至顶网的科技资讯网站，主要面对开源自由软件和关心科技资讯读者群，包括众多中国开源软件的开发者，爱好者和布道者。口号是“奇客的知识，重要的东西”。
NVIDIA Technical Blog发布于 08/21 21:00
Where Security Fits in an AI Agent Stack
（翻译）安全在AI Agent技术栈中的位置
As AI agents become more capable and operate over longer horizons, building security and trust into the applications they power becomes increasingly important.
Google DeepMind Blog发布于 06/25 00:30
Introducing computer use in Gemini 3.5 Flash
（翻译）Gemini 3.5 Flash 引入计算机使用功能
A look at the built-in computer use tool in Gemini 3.5 Flash.
Google DeepMind Blog发布于 06/16 23:46
Securing the future of AI agents
（翻译）保障AI代理的未来
Securing internal systems with an AI Control Roadmap, combining traditional safeguards and real-time monitoring.
Google DeepMind Blog发布于 06/10 18:21
Investing in multi-agent AI safety research
（翻译）投资多智能体AI安全研究
Google DeepMind and partners announce a $10M funding call for multi-agent safety research.
AWS Machine Learning Blog发布于 08/21 00:31
Authoring Dogwood policies from natural language in Amazon Bedrock AgentCore
（翻译）在 Amazon Bedrock AgentCore 中通过自然语言编写 Dogwood 策略
AI agents can take actions that do not match your organization's policies. Policy in Amazon Bedrock AgentCore lets teams enforce controls across agents, now including time-based constraints. This post shows how Policy Authoring turns natural-language policy documents into correct Dogwood policies, w
Hugging Face Blog发布于 07/27 08:00
Anatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the July 2026 Incident
（翻译）前沿实验室智能体入侵剖析：2026年7月事件技术时间线
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
人人都是产品经理发布于 08/26 09:29
边界动作诊断：拒绝 ≠ 终止 权利差异未区分
当AI说“我无法回答”和“会话已结束”时，用户真的能分清吗？本文通过结构化诊断，揭示拒绝与终止在界面上的语义混淆，并引入boundary_action令牌与契约约束，让权利状态清晰可辨，避免用户陷入“被请出门”的困惑。
arXiv cs.AI发布于 08/26 12:00
Auditing the Synthetic Memoir: Measuring Scene-Level Confabulation in LLM-Generated Autobiography Against the Documented Record of the Life It Describes
（翻译）审计合成回忆录：对照其所描述生活的文档记录度量LLM生成自传中的场景级虚构
Abstract page for arXiv paper 2608.23640: Auditing the Synthetic Memoir: Measuring Scene-Level Confabulation in LLM-Generated Autobiography Against the Documented Record of the Life It Describes