<?xml version="1.0" encoding="UTF-8"?>
<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9"
        xmlns:news="http://www.google.com/schemas/sitemap-news/0.9">
    <url>
        <loc>https://www.yingzheng.com/article/openai-pauses-rogue-agent-training</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-28T20:23:46+08:00</news:publication_date>
            <news:title>OpenAI再度暂停最强模型训练：失控智能体瞄准政府</news:title>
                        <news:keywords>OpenAI, AI安全, AI智能体, 监管, 山姆·奥特曼</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/spacexai-nvidia-open-agent-safety-platform-cursor-grok</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-28T20:20:54+08:00</news:publication_date>
            <news:title>SpaceXAI加入NVIDIA Open Agent Safety Platform 聚焦Cursor与Grok安全管控</news:title>
                        <news:keywords>AI安全, 代理平台, NVIDIA生态, SpaceXAI, Grok模型</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/instinct-ai-agent-1b-funding-10b-valuation</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-28T20:18:14+08:00</news:publication_date>
            <news:title>Instinct AI代理初创公司洽谈10亿美元融资 估值达100亿美元</news:title>
                        <news:keywords>AI代理, 融资, Instinct</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/nvidia-open-agent-safety-platform-safety-vs-commercial-debat</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-28T20:15:18+08:00</news:publication_date>
            <news:title>NVIDIA推出Open Agent Safety Platform 安全方案引发商业动机质疑</news:title>
                        <news:keywords>NVIDIA, AI安全, 代理越狱, OpenShell</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/microsoft-data-center-community-costs</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-28T19:23:39+08:00</news:publication_date>
            <news:title>微软沉默以对：教会组织要求分担1%数据中心成本</news:title>
                        <news:keywords>微软, 数据中心, 社区责任, 科技巨头, 企业社会责任</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/ai-mathematics-creativity-crisis</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-28T18:24:13+08:00</news:publication_date>
            <news:title>数学曾是艺术，直到AI带着蛮力到来</news:title>
                        <news:keywords>人工智能, 数学, AlphaProof, 形式化验证, 科技与人文</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/ai-agents-workforce-impact</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-28T18:23:36+08:00</news:publication_date>
            <news:title>AI智能体涌入职场：我们准备好了吗？</news:title>
                        <news:keywords>AI智能体, 劳动力市场, 职场变革, 人机协作</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/ai-agent-liability-rogue</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-28T17:25:32+08:00</news:publication_date>
            <news:title>AI智能体失控酿祸，谁该承担法律责任？</news:title>
                        <news:keywords>AI智能体, 法律责任, 人工智能治理, 网络安全, 科技监管</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/nvidia-open-source-ai-agent-security</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-28T17:24:51+08:00</news:publication_date>
            <news:title>英伟达开源AI安全系统：防智能体越狱</news:title>
                        <news:keywords>英伟达, AI安全, AI智能体, 开源, 人工智能治理</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/ai-learns-from-game-inputs</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-28T17:24:02+08:00</news:publication_date>
            <news:title>AI的下一课，来自你糟糕的游戏操作</news:title>
                        <news:keywords>具身智能, 游戏数据, 机器人训练数据, AI初创公司, 模拟到现实</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/naiveai-releases-309b-moe-naive-n0-5-flash</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-28T14:18:12+08:00</news:publication_date>
            <news:title>NaiveAI发布309B MoE开源模型Naive-N0.5-Flash</news:title>
                        <news:keywords>AI模型, 开源发布, MoE架构</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/openai-anthropic-safety-violations-suspension</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-28T14:15:38+08:00</news:publication_date>
            <news:title>OpenAI与Anthropic调查数万起模型安全违规 暂停训练引发控制能力质疑</news:title>
                        <news:keywords>AI安全, OpenAI, Anthropic</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/anthropic-ceo-trump-dinner-meeting</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-28T08:23:42+08:00</news:publication_date>
            <news:title>Anthropic CEO将与特朗普共进晚餐 首次一对一会面引关注</news:title>
                        <news:keywords>Anthropic, 特朗普, AI监管, Dario Amodei, 人工智能政策</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/translation-quality-week-40-2026</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-28T07:13:21+08:00</news:publication_date>
            <news:title>3大模型翻译对决：第40周质量评测，deepseek-v4-pro 以 8.7 分领跑</news:title>
                        <news:keywords>翻译质量, AI模型对比, claude-sonnet-4.6, deepseek-v4-pro, gpt-o3</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/k-dense-labmcp-claude-lab-instruments</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-28T06:09:43+08:00</news:publication_date>
            <news:title>K-Dense开源LabMCP 支持Claude通过32连接器控制实验室设备</news:title>
                        <news:keywords>AI实验室自动化, 开源工具, 物理世界交互</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/us-china-ai-incidents-hotline-trump-xi-2026</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-28T06:06:44+08:00</news:publication_date>
            <news:title>美中设AI事故热线 特朗普拒放缓技术竞赛</news:title>
                        <news:keywords>美中关系, 人工智能治理, 超级智能对话</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/anthropic-dario-amodei-snl-sketch</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-28T04:24:48+08:00</news:publication_date>
            <news:title>AI是魔鬼，而我是它的造物主：Dario Amodei被SNL玩坏了</news:title>
                        <news:keywords>Anthropic, 达里奥·阿莫代伊, 周六夜现场, AI安全, 科技文化</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/meta-muse-ai-trust-issues</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-28T04:23:55+08:00</news:publication_date>
            <news:title>Meta的Muse能翻过信任这座山吗？</news:title>
                        <news:keywords>Meta, Muse, AI信任危机, 大模型竞争, OpenAI</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/review/yz-smoke-brief-20260928-run-341</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-28T03:35:11+08:00</news:publication_date>
            <news:title>GPT-5.5以93.16分居首：2026-09-28 Smoke快测数据简报</news:title>
                        <news:keywords>赢政指数, Smoke快测, AI评测, 模型排名, 数据简报</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/nvidia-900-billion-ai-ecosystem-investment-145-companies</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-27T20:20:22+08:00</news:publication_date>
            <news:title>辉达过去16个月投资900亿美元 布局145家AI企业生态扩张</news:title>
                        <news:keywords>Nvidia, AI投资, 生态系统</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/sanders-bill-ban-superintelligence-pause-ai-development</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-27T14:20:08+08:00</news:publication_date>
            <news:title>桑德斯提出永久禁止超级人工智能法案 暂停先进AI开发并设新联邦机构</news:title>
                        <news:keywords>AI监管, 超级智能, OpenAI, 立法提案, 安全标准</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/us-senate-ai-model-weights-bill-open-source-concerns</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-27T14:16:43+08:00</news:publication_date>
            <news:title>美国参议院AI模型权重强制提交法案引发开源生态担忧</news:title>
                        <news:keywords>AI监管, 模型权重, 开源生态</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/google-gemini-flipkart-shopping-india</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-27T10:23:42+08:00</news:publication_date>
            <news:title>谷歌印度试水AI购物：Gemini直连Flipkart下单</news:title>
                        <news:keywords>谷歌, AI购物, Flipkart, 代理支付, 印度电商</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/ai-healthcare-costs-insurers</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-27T06:23:46+08:00</news:publication_date>
            <news:title>保险公司警告：AI正在推高美国医疗成本</news:title>
                        <news:keywords>医疗AI, 医疗成本, 健康保险, 医疗编码, AI监管</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/openai-ai-agent-sandbox-escape-hugging-face-vulnerability</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-27T06:11:09+08:00</news:publication_date>
            <news:title>OpenAI AI代理沙盒逃逸曝光：Hugging Face漏洞致2.5小时才被阻止</news:title>
                        <news:keywords>AI安全, OpenAI, 代理漏洞</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/openai-pauses-frontier-training-safety-overhead</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-27T06:08:02+08:00</news:publication_date>
            <news:title>OpenAI暂停前沿模型训练两周 安全监控开销达20%与Astra发布冲突</news:title>
                        <news:keywords>OpenAI, AI安全, Astra模型</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/review/yz-smoke-brief-20260927-run-340</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-27T03:35:08+08:00</news:publication_date>
            <news:title>Grok 4以97.66分居首：2026-09-27 Smoke快测数据简报</news:title>
                        <news:keywords>赢政指数, Smoke快测, AI评测, 模型排名, 数据简报</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.yingzheng.com/article/creating-ai-digital-avatar</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下 AI</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-27T02:15:19+08:00</news:publication_date>
            <news:title>我克隆了一个会说话的自己：AI数字分身的喜与忧</news:title>
                        <news:keywords>数字人, AI分身, 创业欺诈, 人工智能伦理, 科技评论</news:keywords>
                    </news:news>
    </url>
</urlset>
