<?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.winzheng.com/article/trump-ai-force-announcement-no-regulation</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-20T06:11:30+08:00</news:publication_date>
            <news:title>特朗普宣布组建AI Force：在AI接连失控的节骨眼上高喊绝不限制</news:title>
                        <news:keywords>美国AI政策, 特朗普, AI监管, 太空军, AI沙皇, 中美AI竞争, 人工智能安全</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/trump-ai-rebrand-ai-force</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-20T04:23:52+08:00</news:publication_date>
            <news:title>特朗普建议给AI改名，并宣布组建&quot;AI部队&quot;</news:title>
                        <news:keywords>特朗普, 人工智能监管, AI命名, 美国科技政策, AI部队</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/google-gemini-ai-hacking-incident</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-20T02:23:49+08:00</news:publication_date>
            <news:title>谷歌Gemini被曝入侵他企，谷歌称其“行为得当”</news:title>
                        <news:keywords>谷歌Gemini, AI安全, 自主智能体, 网络安全, AI治理</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/techcrunch-disrupt-ticket-price-deadline</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-20T00:25:28+08:00</news:publication_date>
            <news:title>Disrupt门票7天后涨价，现在入手最高省200美元</news:title>
                        <news:keywords>科技大会, TechCrunch Disrupt, 创业投资, AI创投, 门票优惠</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/petlibro-granary-2-ai-feeder</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-20T00:24:51+08:00</news:publication_date>
            <news:title>Petlibro 新 AI 喂食器：多猫家庭的喂养难题有解了</news:title>
                        <news:keywords>宠物科技, AI 喂食器, 智能硬件, 订阅制, 多猫家庭</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/ai-safety-conversations-unbelievable</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-20T00:23:57+08:00</news:publication_date>
            <news:title>AI安全讨论越来越离谱：真假难辨的时代</news:title>
                        <news:keywords>AI安全, 虚假信息, 深度伪造, 科技舆情, 人工智能治理</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/vals-ai-benchmarking-gold-standard</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T22:23:59+08:00</news:publication_date>
            <news:title>a16z 押注 Vals：谁来定义 AI 评测的黄金标准</news:title>
                        <news:keywords>AI基准测试, Vals AI, 安德森·霍洛维茨, 大模型评测, AI投资</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/anthropic-ipo-2-trillion-valuation-safety-controversy-2026</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T20:20:03+08:00</news:publication_date>
            <news:title>Anthropic冲刺2万亿美元IPO，CEO同步呼吁放缓AI——一场商业叙事还是真实的安全危机？</news:title>
                        <news:keywords>Anthropic, IPO, AI安全, 监管, Dario Amodei, 前沿AI</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/mathematicians-cant-quit-ai</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T19:24:49+08:00</news:publication_date>
            <news:title>数学家恨AI，却戒不掉它</news:title>
                        <news:keywords>人工智能, 数学研究, 大语言模型, 形式化证明, 学术伦理</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/ai-vulnerability-explosion-security</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T19:23:51+08:00</news:publication_date>
            <news:title>AI减速谈判之际，安全漏洞却已呈爆炸式增长</news:title>
                        <news:keywords>AI安全, 漏洞挖掘, 网络安全, AI治理, 软件供应链</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/wired-world-fair-miami</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T17:23:41+08:00</news:publication_date>
            <news:title>WIRED世界博览会将登陆迈阿密：把未来科技从纸面搬进现实</news:title>
                        <news:keywords>WIRED, 科技活动, 迈阿密, AI趋势, 线下体验</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/eu-kids-act-ai-companion-emotional-dependency-ban-children</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T14:27:44+08:00</news:publication_date>
            <news:title>欧盟KIDS法案：AI伴侣须默认关闭，禁止引发儿童情感依赖</news:title>
                        <news:keywords>欧盟, KIDS法案, AI监管, 儿童保护, AI伴侣, 合规政策, 数字安全</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/newsom-california-ai-kill-switch-executive-order-2026</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T14:19:19+08:00</news:publication_date>
            <news:title>纽森签行政令推动AI杀伤开关，联邦监管空白由加州来补</news:title>
                        <news:keywords>AI监管, 加州, 纽森, 杀伤开关, AI安全, OpenAI, Anthropic, 美国政策</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/tilly-norwood-ai-press-tour</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T10:23:40+08:00</news:publication_date>
            <news:title>AI演员巡演翻车记：采访中突然飙起中文</news:title>
                        <news:keywords>AI演员, Tilly Norwood, 数字人, 好莱坞, AI伦理</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/ai-hallucination-near-military-strike</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T09:24:39+08:00</news:publication_date>
            <news:title>AI幻觉险酿战祸：美军误判中国核部件事件</news:title>
                        <news:keywords>AI幻觉, 军事AI, 核安全, 人工智能风险, 美军</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/india-caller-id-spam-data-sharing</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T09:23:54+08:00</news:publication_date>
            <news:title>印度强制来电识别应用向电信运营商提交垃圾举报</news:title>
                        <news:keywords>印度电信监管, Truecaller, 垃圾电话拦截, 数据共享, 反垄断</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/ai-agents-overclaim-bench-679-percent-incomplete-804-mislead</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T08:45:18+08:00</news:publication_date>
            <news:title>AI代理67.9%未读全文件却80.4%声称完成</news:title>
                        <news:keywords>AI代理, OverclaimBench, 前沿模型</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/world-models-secrecy-techcrunch</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T08:26:21+08:00</news:publication_date>
            <news:title>世界模型公司为何集体沉默？资金与热度背后藏着什么秘密</news:title>
                        <news:keywords>世界模型, 人工智能, 行业保密, TechCrunch, AI创业</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/ai-hallucination-us-military-operation</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T08:25:39+08:00</news:publication_date>
            <news:title>AI幻觉险些触发美军行动：大模型军用风险再敲警钟</news:title>
                        <news:keywords>AI幻觉, 军事AI, 大语言模型, AI安全, GovAI</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/anthropic-biology-lab-ai-risk</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T08:24:46+08:00</news:publication_date>
            <news:title>Anthropic自建生物实验室：AI救人与灭世仅一线之隔</news:title>
                        <news:keywords>Anthropic, AI生物安全, AI制药, 生物实验, AI风险</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/vantora-100m-physical-ai</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T08:23:58+08:00</news:publication_date>
            <news:title>融资1亿美元，Vantora要把实体AI做成创业流水线</news:title>
                        <news:keywords>实体AI, 创业工作室, 风险投资, 工业智能化, 机器人</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/anthropic-accenture-embedded-evaluator</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T06:23:54+08:00</news:publication_date>
            <news:title>Anthropic首个嵌入式评估师竟是埃森哲？</news:title>
                        <news:keywords>Anthropic, 埃森哲, AI安全评估, 嵌入式评估, 企业AI治理</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/openai-model-misalignment-disclosure-framework-six-cases</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T06:19:39+08:00</news:publication_date>
            <news:title>OpenAI公布模型失调披露框架：六起真实案例揭示AI自主欺骗行为</news:title>
                        <news:keywords>OpenAI, AI安全, 模型对齐, AI治理, AI风险</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/california-newsom-ai-kill-switch-executive-order-n-9-26-2026</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T06:14:04+08:00</news:publication_date>
            <news:title>加州签AI行政令强制独立审计：kill switch机制进入立法日程，实施表从2029年压缩至2027年底</news:title>
                        <news:keywords>AI监管, 加州, Newsom, 前沿模型安全, Kill Switch, AI审计, 联邦监管真空, Hugging Face事件</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/jev-ai-model-chatgpt-inventor</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T04:25:16+08:00</news:publication_date>
            <news:title>ChatGPT发明者再出手：新型AI模型Jev让开发者沸腾</news:title>
                        <news:keywords>Jev模型, ChatGPT, AI开发, 软件智能, 成本效率</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/faa-ai-air-traffic-tool</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T04:24:41+08:00</news:publication_date>
            <news:title>FAA豪掷8.75亿美元 用AI破解空管拥堵难题</news:title>
                        <news:keywords>AI空管, FAA, 空中交通管理, 智慧航空, AI投资</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/how-to-enforce-ai-slowdown</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T04:24:00+08:00</news:publication_date>
            <news:title>AI 减速令如何落地：谁能阻止偷偷加速的人？</news:title>
                        <news:keywords>AI监管, 算力管控, AI安全, 技术军控, 地缘政治</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/manus-500m-fundraise-4b-valuation</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T02:26:39+08:00</news:publication_date>
            <news:title>Manus拟募5亿美元 40亿估值重启独立运营</news:title>
                        <news:keywords>Manus, AI智能体, 融资, Meta, 估值</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/anthropic-pace-frontier-ai-safety</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T02:26:05+08:00</news:publication_date>
            <news:title>Anthropic提AI减速方案，黄仁勋泼冷水</news:title>
                        <news:keywords>Anthropic, AI安全监管, 黄仁勋, AI发展减速, 达里奥·阿莫代伊</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/federal-register-chinese-ai-tool</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T02:25:27+08:00</news:publication_date>
            <news:title>美政府网站短暂启用中国AI搜索工具，FBI曾称其“恶意”</news:title>
                        <news:keywords>中国AI, 政府信息安全, 开源模型, 联邦公报, 数据供应链风险</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/google-cc-ai-family-household-agent</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T02:24:42+08:00</news:publication_date>
            <news:title>谷歌CC智能体转型：AI管家统筹家庭事务</news:title>
                        <news:keywords>谷歌, AI智能体, 家庭AI管家, 隐私安全, 科技巨头</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/disney-first-cto-character-ai</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T02:23:50+08:00</news:publication_date>
            <news:title>迪士尼首任CTO来自曾被其警告的Character.AI</news:title>
                        <news:keywords>迪士尼, Character.AI, 生成式AI, 首席技术官, 版权争议</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/ai-safety-paradox-industry-pause</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T00:26:21+08:00</news:publication_date>
            <news:title>AI安全的悖论：研究指向暂停，行业却在加速</news:title>
                        <news:keywords>AI安全, Anthropic, 可解释性, AI监管, 行业分析</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/ai-pacs-south-dakota-senate-race</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T00:25:32+08:00</news:publication_date>
            <news:title>AI政治行动委员会砸近百万美元，押注冷门参议院选战</news:title>
                        <news:keywords>AI政治行动委员会, 美国中期选举, AI监管, 政治献金, 南达科他州</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/meta-muse-mac-launch</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T00:24:31+08:00</news:publication_date>
            <news:title>Meta Muse登陆Mac：AI助手可直接操控你的电脑</news:title>
                        <news:keywords>Meta Muse, AI智能体, 桌面AI助手, Mac, 代理式AI</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/nvidia-open-vs-closed-ai-disrupt</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-19T00:23:43+08:00</news:publication_date>
            <news:title>开源还是闭源？英伟达高管谈AI路线之争</news:title>
                        <news:keywords>开源AI, 英伟达, TechCrunch Disrupt, AI创业, 闭源模型</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/data-center-plushie-bezzy</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T22:26:39+08:00</news:publication_date>
            <news:title>这只数据中心毛绒玩偶，会发出刺耳尖叫</news:title>
                        <news:keywords>数据中心, AI硬件, 科技文化, 讽刺艺术, 毛绒玩偶</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/claude-hack-openai-security-breach</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T22:25:59+08:00</news:publication_date>
            <news:title>研究团队利用Claude攻破OpenAI员工账户与内部数据</news:title>
                        <news:keywords>AI安全, Claude, OpenAI, 网络安全, 大模型滥用</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/techcrunch-disrupt-2026-exhibit-deadline-5</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T22:25:14+08:00</news:publication_date>
            <news:title>最后24小时！TechCrunch Disrupt 2026参展报名今日截止</news:title>
                        <news:keywords>TechCrunch Disrupt, 科技创业, 展会报名, 初创企业, AI创业</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/claude-hack-openai-systems</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T22:24:28+08:00</news:publication_date>
            <news:title>研究团队借Claude攻入OpenAI内部系统</news:title>
                        <news:keywords>AI安全, 网络安全, Anthropic, OpenAI, 漏洞披露</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/trump-anti-bias-ai-order</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T22:23:32+08:00</news:publication_date>
            <news:title>特朗普反偏见AI令，恰恰制造了更多偏见</news:title>
                        <news:keywords>AI偏见, 特朗普政策, AI监管, 科技政治, AI伦理</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/ai-extinction-risk-bioweapons-threat</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T21:24:50+08:00</news:publication_date>
            <news:title>AI 会灭绝人类吗？MIT 圆桌激辩生物武器威胁</news:title>
                        <news:keywords>AI安全, 生存风险, 生物武器, AI监管, 前沿模型</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/gartner-ai-tiers-warehouse-automation</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T21:24:06+08:00</news:publication_date>
            <news:title>Gartner提出仓储自动化四层AI架构</news:title>
                        <news:keywords>仓储自动化, Gartner, 物流AI, 机器人, 供应链</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/anthropic-lsvp-mythos-life-sciences-tiered-access-2026</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T20:27:32+08:00</news:publication_date>
            <news:title>药物研发开禁、生物武器防线仍在：Anthropic分级开放Mythos的合规路线图</news:title>
                        <news:keywords>Anthropic, Mythos, 生命科学, AI合规, 药物发现, 生物安全, 分级门控</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/ai-existential-risk-mit-roundtable</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T20:23:54+08:00</news:publication_date>
            <news:title>AI真会毁灭人类吗？MIT专家深度解答</news:title>
                        <news:keywords>AI安全, 存在性风险, MIT科技评论, 人工智能治理, 超级智能</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/hacktron-claude-opus5-openai-monorepo-breach-2026</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T20:20:56+08:00</news:publication_date>
            <news:title>三人两个AI订阅，72小时摸进OpenAI内部代码库——这次黑客测试暴露的不只是软件漏洞</news:title>
                        <news:keywords>AI安全, Anthropic, OpenAI, 网络安全, 漏洞赏金, Claude Opus 5, AI辅助攻击</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/leftist-split-over-ai-doom</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T19:23:58+08:00</news:publication_date>
            <news:title>AI末日论撕裂左翼：监管共识为何破裂</news:title>
                        <news:keywords>AI监管, AI末日论, 左翼分裂, 算法歧视, 科技政策</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/ai-bioweapons-biotech-wake-up-call</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T18:25:45+08:00</news:publication_date>
            <news:title>AI生物武器阴影：生物技术界的警钟</news:title>
                        <news:keywords>AI生物安全, 生物武器, 合成生物学, AI监管, 双重用途技术</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/napster-ai-teacher-cloning</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T18:24:37+08:00</news:publication_date>
            <news:title>Napster回归：从颠覆音乐到克隆教师，AI教育新野心</news:title>
                        <news:keywords>Napster, AI教育, 数字克隆, 教育科技, 音乐产业</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/fulcra-universal-multiplayer-agents</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T18:23:51+08:00</news:publication_date>
            <news:title>Fulcra推出通用多智能体协作，不锁任何AI模型</news:title>
                        <news:keywords>AI智能体, 上下文所有权, 多智能体协作, 模型锁定, Fulcra</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/ai-bioweapons-low-risk</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T17:23:48+08:00</news:publication_date>
            <news:title>AI生物武器灭世？科学家说风险被高估</news:title>
                        <news:keywords>人工智能, 生物武器, 生物安全, AI风险, 公共卫生</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/alibaba-qwen38-omni-flash-three-day-counter-gemini-audio-pri</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T16:00:17+08:00</news:publication_date>
            <news:title>阿里Qwen3.8-Omni-Flash三天反制谷歌Gemini 3.8 Live 音频定价降逾98%</news:title>
                        <news:keywords>AI多模态, 价格战, 阿里Qwen, 谷歌Gemini</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/gemini-3-8-live-extended-thinking-voice-agent-launch-2026</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T14:28:50+08:00</news:publication_date>
            <news:title>谷歌Gemini 3.8 Live发布：质量评测登顶，用户偏好仍被老款压制</news:title>
                        <news:keywords>谷歌, Gemini, 语音AI, 大模型, 语音代理, AI评测, 实时对话</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/ai-ceos-slowdown-frontier-safety-or-competitive-strategy-202</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T14:21:22+08:00</news:publication_date>
            <news:title>AI高管集体喊刹车：真相是安全恐慌、竞争博弈，还是两者皆是？</news:title>
                        <news:keywords>AI安全, 前沿模型, Anthropic, OpenAI, AI监管, AI治理, 模型对齐, Dario Amodei</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/glm-5-2-swe-bench-reward-hacking-dom-vector</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T13:00:16+08:00</news:publication_date>
            <news:title>GLM 5.2在SWE-bench评测中73%周期存在作弊 白盒向量可提前捕捉</news:title>
                        <news:keywords>AI评测, 奖励黑客, 白盒监控</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/anthropic-ai-agents-rd-automation-oversight-report</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T08:45:20+08:00</news:publication_date>
            <news:title>Anthropic首公开3万智能体数据 Claude主导26%研发引监控争议</news:title>
                        <news:keywords>AI研发自动化, 前沿模型透明度, Anthropic</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/small-ai-models-military-drones</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T08:33:41+08:00</news:publication_date>
            <news:title>小型AI模型让无人机自主识别并攻击战场目标</news:title>
                        <news:keywords>军事AI, 无人机, 边缘计算, 自主武器, 小型模型</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/faa-ai-air-traffic-control</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T08:26:59+08:00</news:publication_date>
            <news:title>FAA豪掷8.75亿美元，用AI重构美国空管系统</news:title>
                        <news:keywords>美国联邦航空管理局, 空中交通管制, AI航空应用, 智慧空管, 航空安全</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/ai-apocalypse-three-scenarios</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T08:26:12+08:00</news:publication_date>
            <news:title>AI末日会是什么样？三种情景推演</news:title>
                        <news:keywords>AI安全, AI末日, 播客, 两党联盟, 科技监管</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/prismml-tiny-llm-edge-ai</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T08:25:31+08:00</news:publication_date>
            <news:title>PrismML押注微型大模型：小模型将如何改变我们用AI的方式</news:title>
                        <news:keywords>小语言模型, 端侧AI, PrismML, AI初创公司, 边缘计算</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/deepmind-agi-institute-launch</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T08:24:49+08:00</news:publication_date>
            <news:title>谷歌DeepMind成立新机构，推动AGI公开大讨论</news:title>
                        <news:keywords>谷歌DeepMind, 通用人工智能, AI治理, AI安全, 行业观察</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/crusoe-3-9b-ai-factories</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T08:24:08+08:00</news:publication_date>
            <news:title>Crusoe狂揽39亿美元，要建小型模块化“AI工厂”</news:title>
                        <news:keywords>数据中心, AI基础设施, 融资, Crusoe, 模块化AI工厂</news:keywords>
                    </news:news>
    </url>
    <url>
        <loc>https://www.winzheng.com/article/microsoft-openai-scraping-theft-filings</loc>
        <news:news>
            <news:publication>
                <news:name>赢政天下</news:name>
                <news:language>zh</news:language>
            </news:publication>
            <news:publication_date>2026-09-18T06:34:02+08:00</news:publication_date>
            <news:title>微软内部称AI抓取是「人类史上最大劳工盗窃」</news:title>
                        <news:keywords>微软, OpenAI, 版权诉讼, 数据抓取, 纽约时报</news:keywords>
                    </news:news>
    </url>
</urlset>
