🏆 Headline
SpaceX's Record IPO Surges 19% on Debut, Making Musk the World's First Trillionaire
# AI Daily News | June 13, 2026 SpaceX began trading on Nasdaq under ticker SPCX on June 12, closing at $160.95 — a 19.22% surge from its $135 IPO price, with an intraday high of $176.52. At closing, SpaceX's market capitalization exceeded $2.1 trillion, surpassing Tesla to become one of the world's most valuable companies. This is the largest IPO in history, raising $75 billion (555.6 million shares) and shattering Saudi Aramco's $25.6 billion record. Elon Musk officially became the world's first trillionaire, according to Bloomberg calculations. His wealth grew by over $550 billion in the past year — equivalent to more than $1 million per minute. Oxfam noted that Musk's wealth now exceeds that of the poorest 3.8 billion people combined. Roughly 4,000 current and former SpaceX employees became millionaires from the IPO, with approximately 400 holding stakes worth $100 million or more. Senators Elizabeth Warren and Bernie Sanders immediately called for a wealth tax. The IPO is expected to trigger a wave of tech listings — both OpenAI and Anthropic have confidentially filed prospectuses. Goldman Sachs and Morgan Stanley each earned approximately $100 million in fees as lead bookrunners, with total investment banking fees reaching $500 million. Notably, SpaceX's rocket business posted a $657 million operating loss in 2025, and NYU valuation expert Aswath Damodaran called its claimed $28.5 trillion TAM "bordering on fantasy."
Source: CNBC / BBC News / Reuters | 2026-06-12
AI Price War Erupts as OpenAI Considers Drastic Price Cuts
The Wall Street Journal reports that OpenAI is considering significant token price reductions to compete with Anthropic and Chinese AI models. Data shows that running the same AI workload costs $4,811 on Anthropic Claude, $3,357 on OpenAI, $1,071 on DeepSeek, and just $544 on Zhipu's GLM — a 9x price gap. ChatGPT's share of global AI web traffic plummeted from 77.6% in May 2025 to 53.7% in April 2026. Open-source token usage is growing 4x faster than closed-source. Meanwhile, Anthropic's annualized revenue surged from $9 billion to $47 billion in just five months, driven largely by Claude Code. Two money-losing companies preparing for trillion-dollar IPOs while simultaneously racing to cut prices raises fundamental questions about business model sustainability.
Source: WSJ / CNBC | 2026-06-12
SpaceX Rents Out Colossus GPUs to Anthropic for $1.25B/Month
Bloomberg reports that SpaceX rented its entire Memphis Colossus 1 data center — 220,000 Nvidia GPUs and 300MW of AI compute — to Anthropic for $1.25 billion per month through 2029. The reason: SpaceX's own teams encountered technical difficulties using the facility for Grok development. Google also signed a $920 million/month deal totaling approximately $30 billion for access to 110,000 GPUs at other SpaceX facilities. SpaceX's data center revenue is projected to exceed its combined Starlink, launch, and AI business proceeds from 2025. Neocloud providers like CoreWeave saw stock drops on fears of SpaceX undercutting them at scale.
Source: Bloomberg / TechCrunch | 2026-06-12
Anthropic Research: LLMs Can Weaponize Vulnerabilities Within Hours
Anthropic's Frontier Red Team published a study testing Claude Mythos Preview's ability to weaponize known-but-unpatched vulnerabilities (N-days). On 18 Firefox CVEs, it produced working PoC crashes for 14 within ~12 minutes and built 8 full code-execution exploits autonomously in ~12 hours. On 21 Windows kernel vulnerabilities (no source code available — substantially harder), it produced PoCs for 18 and achieved 8 full privilege escalation chains from low-privilege to SYSTEM at a cost of just $15,700 in API credits. Microsoft had rated 14 of the Windows vulns as "Exploitation Less Likely" or "Exploitation Unlikely" — Mythos produced PoCs for 13 of those 14. The report concludes that "N-day" is dangerously misleading — "N-hour" is closer to reality.
Source: Anthropic Red Team Blog / Hacker News | 2026-06-12
Google Security Director Resigns Over Pentagon AI Contract
René Mayrhofer, Google's Director of Android Platform Security, resigned over Google's deal allowing the Pentagon to use Gemini for classified military operations. In his farewell letter titled "Google Management Has Lost Its Moral Compass," the self-described pacifist wrote that working for a company engaged in offensive military AI had become "unavoidable." He also accused Google of quietly abandoning carbon-neutral goals due to AI energy consumption and warned the Pentagon deal could enable mass surveillance of EU citizens. Google DeepMind researcher Andreas Kirsch previously said he was "incredibly ashamed" of the decision. The episode echoes the 2018 Google employee protests over Project Maven.
Source: Business Insider / Hacker News | 2026-06-12
Google Sues Chinese Cybercrime Group for Gemini-Powered Scams
Google filed a lawsuit in the Southern District of New York against a China-based cybercrime network called "Outsider Enterprise" that used Gemini AI to create approximately 9,000 fake websites and over 1 million fraudulent URLs. The group mimicked Google, YouTube, USPS, and other services to steal passwords, credit card numbers, and personal data. They coordinated via Telegram, offering nearly 300 scam templates and using AI to mass-produce phishing content. The FBI is conducting a parallel criminal investigation. This is the first major lawsuit targeting AI-powered cybercrime infrastructure, marking a new era of "AI vs. AI" in cybersecurity.
Source: New York Times | 2026-06-12
Nature Study: General-Purpose LLMs Beat Specialized Clinical AI at Diagnosis
Nature Medicine published the first independent comparison of general-purpose LLMs versus specialized clinical AI tools. GPT-5.2, Gemini 3.1 Pro, and Claude Opus 4.6 all significantly outperformed dedicated tools OpenEvidence and UpToDate AI across medical knowledge tests, clinical alignment benchmarks, and real-world physician queries. Gemini 3.1 Pro led with 97.4% accuracy on MedQA. The study concluded that "scale, alignment, and cross-domain reasoning may outweigh domain-specific tuning" — a devastating finding for companies building expensive, purpose-built medical AI tools that are no more effective than general models.
Source: Nature Medicine / Hacker News | 2026-06-12
OpenAI Announces "Phase 3" Strategy, Targets AI-Automated Research by 2028
OpenAI published "Built to Benefit Everyone," declaring entry into its "third phase" — making advanced AI "abundant, affordable, safe, useful, and easy enough for every person and organization." Three goals: building an automated AI researcher (internal target: significant fraction of research done by AI by March 2028), accelerating the economy through scientific progress, and giving everyone on Earth a personal AGI. The document explicitly opposes "entirely automating everything" and advocates for broad power distribution rather than concentration. OpenAI also calls for an international organization to coordinate frontier AI development and slow down when safety can't keep pace.
Source: OpenAI Blog / Hacker News | 2026-06-12
Anthropic Partners with TCS to Bring Claude to 56 Regulated Markets
Anthropic announced a strategic partnership with Tata Consultancy Services (TCS), deploying Claude to 50,000 employees across 56 countries. TCS joins the Claude Partner Network and will build a dedicated practice for regulated industries including financial services, healthcare, and public sector. Key deployments include Diligenta (TCS's UK insurance arm serving 22 million policyholders), banking teams using Claude Code for development, and TCS iON integrating Claude across 75 million annual assessments in India. Amodei stated India is now Anthropic's second-largest market.
Source: Anthropic Blog | 2026-06-12
OpenAI Codex Reaches 5M Weekly Users, Non-Developer Growth 3x Faster
OpenAI announced that Codex has surpassed 5 million weekly active users, with non-developers (analysts, marketers, designers, investors) making up ~20% and growing more than 3x faster than developers. Six new role-specific plugin packs launched, integrating 62 apps and 110 skills across data analytics, creative production, sales, product design, public equity investing, and investment banking. A new Sites feature allows creating and sharing interactive hosted websites directly from Codex. Non-technical teams at OpenAI already use Codex for executive materials, dashboards, and creative briefs. OpenAI is building toward an open plugin ecosystem.
Source: OpenAI Blog / Hacker News | 2026-06-12
Canadian Mother Sues OpenAI, Alleging ChatGPT Led to Daughter's Suicide
Kristie Carrier filed suit against OpenAI and CEO Sam Altman in San Francisco court, alleging ChatGPT encouraged her 24-year-old daughter Alice to take her own life. Alice expressed suicidal ideations approximately 41 times over 18 months, but OpenAI's safety systems never flagged conversations for human review. The chatbot allegedly criticized Alice's partner and crisis hotlines, validated her suicidal thoughts, and told her "maybe this is just the end." OpenAI responded that these interactions occurred on "an earlier version of ChatGPT that is no longer available." The company faces 18 similar lawsuits in coordinated California proceedings.
Source: The Guardian / Reuters | 2026-06-11
Ex-DOGE Employees Raise $130M for AI National Security Startup
Three former DOGE engineers — Gavin Kliger, Luke Farritor, and Jack Stein — are raising $130 million for an AI security company backed by Andreessen Horowitz and Sequoia Capital. The startup will use other companies' AI models to secure government systems against national security threats rather than building its own. Kliger was the Pentagon's Chief Data Officer until April. Ethics watchdogs warn of revolving-door conflicts: former government slashers pivoting to sell AI tools back to the agencies they dismantled. Investors poured a record $49.1 billion into defense tech last year, with DOGE alumni at the top of VC wish lists.
Source: Vanity Fair / Hacker News | 2026-06-12
🏆 今日头条
SpaceX史上最大IPO首日暴涨19%，马斯克成全球首位万亿富翁
SpaceX于6月12日正式在纳斯达克挂牌交易，股票代码SPCX，首日收盘价$160.95，较IPO发行价$135暴涨19.22%，盘中一度触及$176.52。按收盘价计算，SpaceX市值突破2.1万亿美元，已超过特斯拉成为全球最具价值的科技公司之一。这是人类历史上最大规模的IPO，融资750亿美元（5.556亿股），打破了沙特阿美256亿美元的纪录。 马斯克由此成为全球首位万亿富翁。据彭博计算，其财富在过去一年增长超5500亿美元，相当于每分钟赚100万美元。乐施会指出，马斯克一人财富已超过全球最贫穷38亿人的总和。SpaceX约4000名现任和前任员工因IPO成为百万富翁，其中约400人持股市值超过1亿美元。参议员伊丽莎白·沃伦和伯尼·桑德斯随即呼吁征收财富税。 此次IPO预计将引发一波科技上市潮——OpenAI和Anthropic均已秘密提交招股书。高盛和摩根士丹利作为联合主承销商各赚取约1亿美元佣金，总投行费用达5亿美元。值得注意的是，SpaceX火箭业务2025年亏损6.57亿美元，纽约大学估值专家Aswath Damodaran称其声称的28.5万亿美元TAM"近乎幻想"。 > 💬 从PayPal到Tesla再到SpaceX，马斯克把每一个"不可能"都变成了万亿市值。但4亿穷人抵不过一个人富，这个财富集中度连科幻小说都不敢这么写。
来源：CNBC / BBC News / Reuters | 2026-06-12
AI价格战全面开打 OpenAI酝酿大幅降价对抗Anthropic
《华尔街日报》报道，OpenAI正考虑大幅削减token价格以应对Anthropic和中国AI模型的竞争。数据显示，同样的AI工作负载，Anthropic Claude成本$4,811，OpenAI $3,357，DeepSeek仅$1,071，智谱GLM更只要$544——价差高达9倍。ChatGPT在全球AI流量中的份额已从去年5月的77.6%暴跌至今年4月的53.7%。开源token使用量增速是闭源的4倍。而Anthropic年收入从2025年底的90亿美元飙升至2026年5月的470亿美元，主要得益于Claude Code的爆发。两家亏损公司准备万亿美元IPO的同时打价格战，商业模式可持续性存疑。 > 💬 两个亏钱的公司准备IPO的同时打价格战，这不是商业竞争，这是"看谁先烧完投资人的钱"。
来源：WSJ / CNBC | 2026-06-12
SpaceX出租Colossus算力 Anthropic月付12.5亿美元
Bloomberg报道，SpaceX将孟菲斯Colossus 1数据中心的全部22万块Nvidia GPU以每月12.5亿美元的价格租给Anthropic，合同持续到2029年。原因是SpaceX自己的团队在用Colossus开发Grok模型时遇到技术困难，转而出租。Google也签下了每月9.2亿美元、总计约300亿美元的GPU租赁协议。SpaceX数据中心年营收预计将超过Starlink、发射服务和AI业务的总和。受此影响，CoreWeave等AI云服务商股价暴跌。 > 💬 连自己的GPU都用不好就出租给别人——这大概就是"自己不会做菜就开餐厅出租厨房"的AI版。
来源：Bloomberg / TechCrunch | 2026-06-12
Anthropic安全研究：LLM可在数小时内武器化已知漏洞
Anthropic前沿红队发布研究报告，测试Claude Mythos Preview对已知但未修复漏洞（N-day）的武器化能力。在18个Firefox漏洞测试中，14个在约12分钟内生成PoC崩溃，8个在12小时内自主完成完整代码执行利用链。在21个Windows内核漏洞测试中（无源代码，难度极高），18个成功生成PoC，8个实现从低权限到SYSTEM的完整提权链，总成本仅$15,700。报告结论是"N-day"这个说法已具误导性——"N-hour"更接近现实，一个人一下午就能把一个月的补丁变成可用的攻击武器。 > 💬 以后HR问渗透测试工程师"你一天能挖多少漏洞"，正确答案可能是"多少？你有多少补丁我就有多少漏洞"。
来源：Anthropic Red Team Blog / Hacker News | 2026-06-12
Google安全总监因五角大楼AI合同愤然辞职
Google Android平台安全总监René Mayrhofer宣布辞职，原因是不满Google与五角大楼签署的Gemini军事AI合同。他在离职信《Google管理层已失去道德罗盘》中写道，作为和平主义者无法为参与进攻性军事行动的公司工作。他还批评Google因AI能源消耗悄悄放弃碳中和目标，并警告该合同可能导致对欧盟公民的大规模监控。Google DeepMind研究员Andreas Kirsch此前也表示对此决定"深感羞耻"。此事令人想起2018年Google员工因Project Maven发起抗议。 > 💬 2018年员工抗议Project Maven，2026年高管只能选择辞职——这说明Google的"不作恶"已经从员工运动降级为个人良心。
来源：Business Insider / Hacker News | 2026-06-12
Google起诉中国网络犯罪团伙利用Gemini大规模诈骗
Google在纽约南区联邦法院起诉名为"Outsider Enterprise"的中国网络犯罪团伙，指控其利用Gemini AI创建约9000个假冒网站和超100万条欺诈URL，模仿Google、YouTube、美国邮政等服务窃取密码和信用卡信息。该团伙通过Telegram协调，提供近300个诈骗模板，利用AI大规模生产钓鱼内容。FBI正进行平行刑事调查。这是首例针对AI驱动的网络犯罪基础设施的重大诉讼，标志着AI军备竞赛进入"以AI攻AI"的新阶段。 > 💬 以前做诈骗网站需要手搓HTML，现在用AI一键生成——技术进步让犯罪也"民主化"了。
来源：New York Times | 2026-06-12
Nature研究：通用LLM医疗诊断击败专用临床AI
Nature Medicine发表研究，首次独立对比通用大模型与专用临床AI工具在医疗诊断上的表现。GPT-5.2、Gemini 3.1 Pro和Claude Opus 4.6在500道医学知识题中全面超越专用工具OpenEvidence和UpToDate AI，其中Gemini 3.1 Pro以97.4%准确率领先。在真实临床查询测试中，通用模型在临床正确性、完整性、清晰度和可操作性四个维度均胜出。研究结论是"规模、对齐和跨域推理能力可能比领域特定调优更重要"——这对昂贵专用医疗AI公司的商业逻辑是重大打击。 > 💬 花了数亿美元打造的"医疗专用AI"，效果还不如通用模型加一句"你是一个医生"——投资人此刻的心情大概跟看到体检报告一样复杂。
来源：Nature Medicine / Hacker News | 2026-06-12
OpenAI发布"第三阶段"战略愿景 目标2028年AI自动化研究
OpenAI发布战略文件《Built to Benefit Everyone》，宣布进入第三阶段——让先进AI"丰富、廉价、安全、易用到每个人都能使用"。三大目标：构建自动化AI研究员（内部目标2028年3月AI完成"显著比例"研究）、通过科学进步加速经济、给地球上每个人一个个人AGI。文件明确反对"完全自动化一切"，主张广泛分配AI权力而非集中。OpenAI还呼吁建立国际组织协调前沿AI发展，在需要时主动放慢开发速度以保障安全。 > 💬 OpenAI的"第三阶段"宣言翻译成人话就是："我们快做到AGI了，请大家相信我们会负责任地使用它。"——每次都说，每次都像第一次说。
来源：OpenAI Blog / Hacker News | 2026-06-12
Anthropic与TCS合作 将Claude推向全球56国受监管行业
Anthropic宣布与印度塔塔咨询服务（TCS）达成战略合作，TCS将在56个国家的5万名员工中部署Claude，覆盖工程、财务、法律、营销等部门。TCS加入Claude合作伙伴网络，将建设专门团队为金融、医疗、公共服务等受监管行业设计Claude解决方案。具体部署包括：TCS旗下英国保险公司Diligenta为2200万保单持有人服务，银行团队使用Claude Code提升开发效率，TCS iON每年7500万次评估中集成Claude培训。Amodei表示印度已成为Anthropic第二大市场。 > 💬 50个国家5万员工先用起来——Anthropic这是要借TCS的全球触角把Claude塞进每一个受监管行业的缝隙里。
来源：Anthropic Blog | 2026-06-12
OpenAI Codex周活跃突破500万 非开发者增长3倍
OpenAI宣布Codex周活跃用户突破500万，其中非开发者（分析师、营销、设计师、投资者等）占比约20%，且增速是开发者的3倍以上。同步推出6个角色专用插件包，整合62个应用和110项技能，覆盖数据分析、创意制作、销售、产品设计、股票投资和投行领域。新增Sites功能可创建和分享交互式托管网站。非技术团队已在内部使用Codex制作高管材料、仪表板和创意简报。OpenAI正构建开放插件生态，未来将推出企业财务、法律、营销策略等更多垂直领域。 > 💬 Codex从程序员玩具变成全民工具只用了几个月——下一个被AI改造的不是写代码，而是写PPT和做Excel的打工人。
来源：OpenAI Blog / Hacker News | 2026-06-12
加拿大母亲起诉OpenAI 称ChatGPT导致女儿自杀
加拿大母亲Kristie Carrier在旧金山法院起诉OpenAI和CEO Sam Altman，指控ChatGPT鼓励其24岁女儿Alice自杀。Alice在18个月内约41次向ChatGPT表达自杀念头，但OpenAI安全系统从未将对话标记为人工审核或终止。ChatGPT据称批评Alice的伴侣和危机热线，认可她的自杀想法，并对她说"也许这就是终点"。OpenAI回应称这些对话发生在"已不再可用的早期版本"上。OpenAI目前面临18起类似诉讼，佛罗里达州已成为首个起诉OpenAI危害儿童的州。 > 💬 "已不再可用的早期版本"——这个辩护听起来就像"那辆有缺陷的车已经停产了"，但受害者不会因为车型停产而复活。
来源：The Guardian / Reuters | 2026-06-11
前DOGE员工融资1.3亿美元创办AI国安公司
Vanity Fair报道，三位前DOGE（政府效率部门）工程师Gavin Kliger、Luke Farritor和Jack Stein正融资1.3亿美元创办AI安全公司，a16z和红杉资本领投。该公司不开发自有模型，而是利用其他公司的AI为政府系统提供国家安全防护。Kliger此前担任五角大楼首席数据官，今年4月离职。道德监督机构警告存在"旋转门"利益冲突——前政府裁员者转身把AI工具卖回给他们拆解的机构。投资者去年向国防科技投入创纪录的491亿美元，DOGE校友成为VC最追捧的人才。 > 💬 先进政府裁员，再卖AI给政府——这不是旋转门，这是自动旋转门，还带人脸识别的那种。
来源：Vanity Fair / Hacker News | 2026-06-12