VOL.2026.06.28·31 篇·每日精选
AI 日报
AI DAILY二〇二六年六月二十八日 星期日每日自动更新
今日要点
AI 动态
—Wayfinder Router实现本地与托管LLM间查询的确定性路由，提升模型调用效率。
—OpenAI内部揭秘新模型Sol、Terra和Luna，GPT-5.6或带来多模态与推理能力跃升。
—Anthropic指控阿里巴巴使用2.5万个账户窃取Claude模型，AI竞赛格局重置。
人力资源动态
—Mercer调查显示99%高管预计AI将在两年内引发裁员，劳动力市场面临结构性冲击。
01AI 动态今日动态30 篇
Wayfinder Router：本地与托管LLM之间查询的确定性路由
Hacker NewsWayfinder Router: deterministic routing of queries between local and hosted LLM
GPT-5.6内部揭秘：OpenAI最新AI模型Sol、Terra和Luna带来了什么
Google AI NewsInside GPT-5.6: What OpenAI's newest AI models Sol, Terra and Luna bring to the table
福特雇佣AI并解雇人类，结果适得其反
Hacker NewsFord hired AI and sacked humans. It backfired badly
众议员莫兰提交法案要求进行AI安全报告
Google AI NewsRep. Moran files bill that would require AI safety reporting
对AI垃圾内容的回应来自罗宾·威廉姆斯
Hacker NewsResponse to AI slop is from Robin Williams
中国在网络安全领域已赶上Anthropic，重置AI竞赛格局
Google AI NewsChina Has Matched Anthropic in Cybersecurity, Resetting AI Race
人工智能不会取代银行家——它将重新定义国际银行业：作者Luigi Wewege
Google AI NewsArtificial Intelligence Won't Replace Bankers - It Will Redefine International Banking: By Luigi Wewege
Show HN：KV-psi，利用Linux PSI修剪LLM的KV缓存
Hacker NewsShow HN: KV-psi, using Linux PSI to to trim an LLM KV cache
你所熟知的AI行业今日已死
Hacker NewsThe AI Industry as You Know It Died Today
《小猪佩奇》工作室拟利用AI无限克隆儿童演员声音
Hacker NewsPeppa Pig studio wants to clone child actors' voices with AI indefinitely
FTI Consulting新任AI风险主管能否巧妙重塑FCN的监管与韧性叙事？
Google AI NewsCan FTI Consulting's New AI Risk Leader Subtly Reframe FCN's Regulation and Resilience Narrative?
WEHCO加入针对OpenAI和微软的诉讼
Google AI NewsWEHCO joins lawsuit against OpenAI, Microsoft
人人都曾担心AI接管一切；真正的危险是AI只为少数人服务
Hacker NewsEveryone feared AI taking over; the real danger is AI serving just the few
Show HN: Adrafinil – 仅在代理工作时保持合盖 Mac 唤醒
Hacker NewsShow HN: Adrafinil – keep a lid-closed Mac awake only while agents work
新评论呼吁医疗系统中以患者为中心的 AI 监管
Google AI NewsNew commentary urges patient-centered AI regulation in healthcare systems
New commentary urges patient-centered AI regulation in healthcare systems News-Medical
安全预警系统，看不见的滴滴「基建」
雷锋网安全预警系统，看不见的滴滴「基建」
近日，滴滴在成都办了一场安全开放日，多个司机、乘客、行业专家和媒体被邀请走进滴滴安全预警中心，现场体验一线安全专家如何听音、研判和干预订单风险。这也是滴滴少有地把网约车安全后台摊开给外界看。滴滴创始人、CEO、集团安委会主任程维在活动中表示，保障每天数千万订单的安全，是滴滴身上的责任。滴滴安全体系已经迭代多次，但仍然“还在路上”。他希望通过开放日，让外界看到滴滴如何做安全，也给平台提出更多意见。网约车安全和常规互联网推荐系统不同。推荐、搜索算法通常要在准确率和召回率之间取得平衡，但安全算法不一样。滴滴网约车技术和安全负责人曹乐表示，安全风险是低频事件，但只要平台有能力召回，就不能为了提高准确率而漏掉风险。这意味着，滴滴安全系统必须优先保证召回率。然而，代价是大量误判。据曹乐介绍，
Anthropic 称阿里巴巴使用 2.5 万个账户窃取 Claude 模型
Hacker NewsAnthropic says Alibaba used 25k accounts to mine Claude
2026年6月22日至27日，多达18家来自不同领域的印度初创公司从投资者处融资超10.8亿美元
Google AI NewsBetween June 22 and June 27, 2026, as many as 18 Indian startups from diverse sectors raised over $1.08 billion in funding from investors. These sectors include Proptech, Ecommerce, Martech, Fintech, Spacetech, Cybersecurity, DevOps, Agritech, Logistics,
Between June 22 and June 27, 2026, as many as 18 Indian startups from diverse sectors raised over $1.08 billion in funding from investors. These sectors include Proptech, Ecommerce, Martech, Fintech, Spacetech, Cybersecurity, DevOps, Agritech, Logistics, instagram.com
AI开发者和捐赠者承诺为政治竞选投入2.75亿美元，监管斗争加剧
Google AI NewsAI developers and donors pledge $275M for political races as regulation battles intensify
科技行业应对特朗普在AI政策上的反复无常
Google AI NewsTech industry grapples with Trump’s AI about-faces
Anthropic的Mythos 5获得美国限制措施的有限豁免
Google AI NewsAnthropic's Mythos 5 gets a limited carveout from US restrictions
vivo「再造」折叠屏
雷锋网vivo“再造”折叠屏
手机行业从来不缺挑战。自从2023年中国科技产业开启了所谓的「大模型狂飙时代」后，从互联网巨头到传统制造业，从手机厂商到汽车新势力，中国科技行业迅速陷入了一场前所未有的焦虑当中。在行业落地方向尚未明确时，各行各业均选择了「先上车后补票」的方式乘上这趟「快车」。手机作为最接近消费者的入口，同样也成为了各行各业眼中的「必争之地」。手机厂商如何在激烈的竞争中占据领先地位，早已成为了新的课题。但对手机厂商而言，竞争是立体的，既是向内，也是向外。向内，手机厂商之间的竞争已进入消耗战。除了卷参数、卷配置、卷价格三条路径，各家同样开始了AI时代的内卷，大量投入研发资源，争取夺下「AI手机」的心智高地。向外，OpenAI正以65亿美元收购苹果前首席设计官Jony Ive创立的io Products
Anthropic 获准向 100 多家美国组织发布 Claude Mythos 5
Google AI NewsAnthropic Granted Approval To Release Claude Mythos 5 To 100+ U.S. Organizations
Anthropic Granted Approval To Release Claude Mythos 5 To 100+ U.S. Organizations LinkedIn
亚洲AI初创公司推出类Mythos模型，Anthropic出口禁令持续发酵
Google AI NewsAsian AI startups launch Mythos-like models as Anthropic’s export ban drags on
OpenAI将新模型的有限发布限制在美国境内
Google AI NewsOpenAI Restricts Limited Release of New Model to US Only
想要 OpenAI 的最新 AI？你得等等，白宫称 GPT-5.6 将优先供美国使用
Google AI NewsWant OpenAI’s newest AI? You’ll have to wait as White House says GPT-5.6 is for US first
Want OpenAI’s newest AI? You’ll have to wait as White House says GPT-5.6 is for US first Malay Mail
商务部放宽对 Anthropic 的 Mythos 限制，但主要问题仍存
Google AI NewsCommerce Eased Its Block on Anthropic's Mythos, But Major Questions Remain
Commerce Eased Its Block on Anthropic's Mythos, But Major Questions Remain Tech Policy Press
FTC 批准马斯克收购 SpaceX 校友创企 Mesh
Google AI NewsFTC gives Musk the OK to acquire SpaceX alumni startup Mesh
特朗普政府允许Anthropic向特定美国组织发布Mythos
Google AI NewsTrump Administration Allows Anthropic to Release Mythos to Select US Organizations
Chattanooga Times Free Press 出版商加入针对 OpenAI 和微软的诉讼
Google AI NewsChattanooga Times Free Press publisher joins lawsuit against OpenAI, Microsoft
Chattanooga Times Free Press publisher joins lawsuit against OpenAI, Microsoft Chattanooga Times Free Press
03人力资源动态职场与人才1 篇