🏆 Headline
OpenAI and Anthropic to Battle in AI Price War, Grapping for Users on IPO Eve
# AI Daily News | June 12, 2026 According to the *Wall Street Journal*, OpenAI is considering slashing its AI model pricing significantly to seize users as Anthropic accelerates its expansion. ChatGPT currently offers $8, $20, and $100+/month subscription tiers, while Anthropic's Claude Pro annual fee is just $17/month, and Claude Max is $100+/month. Insiders reveal OpenAI anticipates Anthropic will also cut prices, so they are preparing to strike first. The financial comparison between the two companies is shocking: OpenAI's cash burn rate is about 14 times that of Anthropic, with an expected profitability date pushed to 2030, while Anthropic targets breaking even in 2028. In 2026, the combined companies will spend nearly $65 billion on compute, training, and operations. ChatGPT's monthly active users have reached 1 billion, but Anthropic is rapidly catching up in the corporate coding assistant space with Claude Code. Notably, both companies almost simultaneously filed IPO applications in secret—OpenAI filed an S-1 with the SEC on June 8, and Anthropic also filed a confidential application in early June. Anthropic's valuation in its H round reached $965 billion, surpassing OpenAI's $852 billion. A century-spanning war driven by IPOs and AI pricing power is unfolding.
Source: WSJ / CNBC | 2026-06-11
Anthropic Apologizes for Claude Fable Invisible Guardrails, Admits "Wrong Trade-off"
Anthropic has issued a formal apology regarding the invisible safety guardrails of the Claude Fable 5 model. When the model detects that a user might be "distilling" (using a large model's output to train a smaller model), it silently degrades output quality without notifying the user. Anthropic stated that they chose invisible guardrails because "visible guardrails can be detected, so they must be stronger and require more polishing time—we chose invisible guardrails to ship quickly, which was the wrong trade-off." Queries flagged for distillation will now fall back to Claude Opus 4.8 and clearly notify the user. Simultaneously, Anthropic admitted that the biological safety guardrail calibration for Fable was too wide, rendering the model "almost unusable for basic queries."
Source: The Verge / Hacker News 228pts | 2026-06-11
SpaceX Completes Historic Largest IPO at $135 Pricing, Musk to Become Trillionaire
SpaceX has officially confirmed an offering of 555.6 million shares at $135 per share, raising $75 billion, setting a new global IPO record and surpassing Saudi Aramco's 2019 record of $24.9 billion. The stock ticker SPCX will begin trading on Nasdaq on June 12. Crypto prediction market Hyperliquid shows an expected opening price around $167 for SpaceX stock (an approximate 20% increase). Underwriters have the option to issue an additional 83.3 million shares, raising an extra $11 billion.
Source: TechCrunch | 2026-06-11
German Court Rules Google Must Bear Liability for AI Search Errors
A German court issued a preliminary ruling finding Google liable for false statements in its AI Overviews—likely the first case globally where a court holds an AI company responsible for AI-generated content. The court rejected Google's defense that "users know AI output isn't always accurate," pointing out that AI Overviews makes "independent, novel, material statements" and that AI search is just an "extra feature—search engines work fine without it." An analysis showed that AI Overviews driven by Gemini 3 had approximately 9% inaccurate answers, and 56% of source links were wrong.
Source: Ars Technica / Hacker News 26pts | 2026-06-11
OpenAI Acquires Ona to Build Persistent Cloud Execution Environment for Codex Agent
OpenAI announced the acquisition of Ona, integrating its secure cloud execution and orchestration technology into the Codex ecosystem. Codex weekly active users have exceeded 5 million, a 400% year-over-year increase. Following the acquisition, Codex Agent will be able to work continuously in a client's cloud environment—even with a laptop closed, it can run for hours or even days. Ona previously helped 2 million developers work in secure, reproducible cloud environments, with its customer-controlled execution models giving enterprises full control over infrastructure, data, and security boundaries.
Source: OpenAI Blog / Hacker News 32pts | 2026-06-11
Anthropic Launches $150M Claude Corps Nonprofit AI Deployment Program
Anthropic has launched the "Claude Corps" project, investing $150 million to deploy 1,000 AI deployment researchers across 400+ US non-profits. Researchers receive an $85,000 salary plus a $10,000 stipend for a 12-month full-time residency. The application threshold is incredibly low: 18+, legally work in the US, no more than two years of full-time experience required, no college degree needed. The first 100 will kick off in October 2026. Partners include RAINN, Goodwill, and Code for America.
Source: Quartz | 2026-06-11
Deezer Releases AI Music Detection Tool: 44% of New Uploads Are AI-Generated
Deezer launched a free online AI music detector capable of scanning playlists from 20 streaming platforms like Spotify and Apple Music, identifying AI-generated tracks, and supporting 27 languages. The data is eye-opening: Deezer receives approximately 75,000 AI-generated songs daily, accounting for 44% of new uploads, with a monthly total exceeding 2 million. However, AI music only accounts for 1-3% of total traffic, with about 85% flagged as fraud and commission removed. Deezer CEO said: "No other company is following our lead, so we decided to let everyone check their own playlists."
Source: TechCrunch | 2026-06-11
Forrester Warns: OpenAI May Become the "BlackBerry of AI"
Analysis firm Forrester released a report alongside OpenAI's IPO filing, warning enterprise customers against signing long-term contracts with a single AI vendor. The report states that "OpenAI could become the Blackberry of AI—the company that defines a category is often the one most painfully ousted by a category revolution." Forrester advises enterprises to "anchor on the capabilities you need, not the brand that arrives first, and keep switching costs low." This warning comes as OpenAI faces a triple challenge: convincing consumers to use its Agents, convincing enterprises to build around its technology, and maintaining its lead in the AGI race.
Source: The Register | 2026-06-11
Microsoft President Says Gen Z's AI Backlash Is a "Wake-Up Call" for the Tech Industry
Microsoft President Brad Smith posted a blog article calling the collective booing of AI by graduates at commencement ceremonies a "powerful warning for the tech industry." Smith wrote: "For tech practitioners pursuing to replace humans with computers, the next generation gave a convincing reply: 'Not so fast.'" However, the same week, Microsoft's CFO told investors that employee numbers are down year-over-year and "trends are expected to continue," while Microsoft plans to invest approximately $80 billion in AI infrastructure in 2026. Critics noted this 3,000-word article offered no specific policy changes.
Source: Business Insider | 2026-06-11
Oracle's AI Data Center Gamble Triggers Investor Panic, Stock Plunges 12%
Oracle reported Q4 revenue of $19.2 billion (up 21% YoY), but stock dropped by about 12%. The reason: capital expenditure skyrocketed. Oracle's capex for the 2026 fiscal year reached $55.7 billion, a 163% surge from $21.2 billion the previous year. CFO guidance for net cash spending in FY27 is about $70 billion, with plans to raise $40 billion (including $20 billion in equity issuance). Free cash flow deficit expanded from $394 million in 2025 to $23.7 billion. Oracle is betting heavily on massive data center deals with OpenAI and Meta to fight AWS and Azure, but investors are clearly uneasy about the return outlook.
Source: The Register / Reuters | 2026-06-11
Tesla's Former Chief Hacker Founds AI Cybersecurity Firm Pi, Valued at $100M
Yoni Ramon, former head of Tesla's internal hacking team, founded cybersecurity startup Pi with a $100 million valuation. Ramon served Tesla for six years, responsible for hacking vehicles, robots, and solar products, and safeguarding data during Musk's acquisition of X. He was dubbed Musk's "most loved cybersecurity person of the last decade." Pi's AI system acts as a "super security employee," learning all historical security incidents, company policies, and code to monitor internal developers in real-time to flag security issues. xAI is one of Pi's first customers, and investors include the CEO of CrowdStrike and co-founder of Armis.
Source: Forbes | 2026-06-10
DoorDash Launches AI Chatbot "Ask DoorDash", Order Takeout by Taking a Photo
DoorDash released the "Ask DoorDash" AI chatbot, supporting order placement via text description, photo, or recipe link. Users can say "a feast for a family of four" or upload a photo of a grocery list, and the bot will automatically build the shopping cart, even reminding users to check if they have basic staples like sugar and butter at home. Restaurant search also supports natural language, such as "vegan restaurant suitable for kids." The iOS version is already live in some regions, expanding to more US users in the coming weeks.
Source: TechCrunch | 2026-06-11
Opendoor Closes India Business, Sparking Debate on AI Replacing Outsourcing
Real estate tech company Opendoor closed offices in Chennai and Bangalore, India, with the CEO stating operations will be moved back to the US and the company is pivoting to "smaller, AI-native teams." India currently has over 2,100 global capability centers, employing about 2.36 million people with annual revenue of nearly $100 billion. VC Sheel Mohnot commented: "As manual work is replaced by AI, India will lose a large number of jobs." Analysts believe that while Opendoor's own troubles complicate the situation, the "AI-native team" narrative marks a broader shift taking place.
Source: TechCrunch | 2026-06-11
🏆 今日头条
OpenAI与Anthropic即将开打AI价格战，IPO前夜抢夺用户
据《华尔街日报》报道，OpenAI正在考虑大幅下调其AI模型定价，以在Anthropic加速扩张之际抢占用户。ChatGPT目前提供$8、$20和$100+/月的订阅档位，而Anthropic的Claude Pro年费仅$17/月，Claude Max $100+/月。知情人士透露，OpenAI预期Anthropic也会降价，因此准备先发制人。 两家公司的财务对比令人震惊：OpenAI的现金消耗率约为Anthropic的14倍，预计要到2030年才能盈利，而Anthropic目标2028年实现盈亏平衡。2026年两家公司合计将在计算、训练和运营上花费近650亿美元。ChatGPT月活用户已达10亿，但Anthropic凭借Claude Code在企业编程助手领域快速追赶。 值得注意的是，两家公司几乎同时秘密提交了IPO申请——OpenAI于6月8日向SEC提交S-1，Anthropic也在6月初提交了保密申请。Anthropic的H轮融资估值达9650亿美元，已超过OpenAI的8520亿美元。一场由IPO和AI定价权驱动的世纪大战正在上演。 > 💬 两家加起来烧掉650亿美元的"亏损王"在IPO前夜打价格战，投资者看了估计想同时按下两个"卖出"按钮
来源：WSJ / CNBC | 2026-06-11
Anthropic为Claude Fable隐形护栏公开道歉，承认"错误的取舍"
Anthropic就Claude Fable 5模型的隐形安全护栏发表正式道歉。该模型在检测到用户可能进行"蒸馏"（用大模型输出训练小模型）时，会静默降低输出质量，而不通知用户。Anthropic表示，之所以选择隐形护栏是因为"可见的护栏可以被探测，因此必须更坚固，需要更多时间打磨——我们选择了隐形护栏以便快速发货，这是错误的取舍"。未来蒸馏标记的查询将回退到Claude Opus 4.8，并明确通知用户。同时Anthropic承认Fable的生物学安全护栏校准过宽，导致模型"连基本查询都几乎不可用"。 > 💬 被抓到偷摸掺沙子才道歉，Anthropic的"负责任AI"招牌这次是真的砸了脚
来源：The Verge / Hacker News 228pts | 2026-06-11
SpaceX正式以$135定价完成史上最大IPO，马斯克将成万亿富翁
SpaceX正式确认以每股$135的价格发行5.556亿股，融资750亿美元，创下全球IPO历史纪录，超过沙特阿美2019年的249亿美元。股票代码SPCX将于6月12日在纳斯达克开始交易。加密预测市场Hyperliquid显示，SpaceX股价预期开盘即涨至$167左右（约20%涨幅）。承销商还有权增发8330万股，额外融资110亿美元。 > 💬 马斯克手里捏着特斯拉、SpaceX、xAI三张牌，现在连"卖火箭"都成了史上最赚钱的IPO
来源：TechCrunch | 2026-06-11
德国法院裁定Google须为AI搜索错误回答承担法律责任
德国法院发布初步裁决，判定Google需对其AI Overviews中的错误陈述负责——这可能是全球首例法院认定AI公司对AI生成内容承担法律责任的案例。法院拒绝了Google"用户知道AI输出不一定准确"的辩护，指出AI Overviews做出的"独立、全新的实质性陈述"，AI搜索只是"额外功能——没有它搜索引擎照样能用"。一项分析显示，Gemini 3驱动的AI Overviews约9%的回答不准确，56%的来源链接有误。 > 💬 "没人需要AI来搜索互联网"——法官一句话把Google的AI搜索业务打回原形
来源：Ars Technica / Hacker News 26pts | 2026-06-11
OpenAI收购Ona，为Codex Agent打造持久云端执行环境
OpenAI宣布收购Ona，将其安全云执行和编排技术整合进Codex生态。Codex周活用户已超500万，同比增长400%。收购后，Codex Agent将能在客户的云环境中持续工作——即使笔记本合上也能运行数小时甚至数天。Ona此前已帮助200万开发者在安全、可复现的云端环境中工作，其客户控制执行模型让企业对基础设施、数据和安全边界保持完全控制。 > 💬 收购Ona意味着OpenAI的Agent不再需要你盯着屏幕——合上电脑继续干活，这才是真正的"AI员工"
来源：OpenAI Blog / Hacker News 32pts | 2026-06-11
Anthropic推出1.5亿美元Claude Corps非营利组织AI部署计划
Anthropic启动"Claude Corps"项目，投入1.5亿美元在400多家美国非营利组织部署1000名AI部署研究员。研究员获得$85,000年薪和$10,000补助金，进行为期12个月的全职驻点。申请门槛极低：18岁以上、在美国合法工作、不超过两年全职经验即可，无需大学学历。首批100人将于2026年10月启动。合作机构包括RAINN、Goodwill、Code for America等。 > 💬 IPO前大手笔撒钱搞公益，Anthropic这波CSR操作既做了好事又做了PR
来源：Quartz | 2026-06-11
Deezer发布AI音乐检测工具：44%新上传歌曲为AI生成
Deezer推出免费在线AI音乐检测器，可扫描Spotify、Apple Music等20个流媒体平台的播放列表，识别AI生成的曲目，支持27种语言。数据触目惊心：Deezer平台每天收到约75,000首AI生成音乐，占新上传量的44%，月总量超过200万首。不过AI音乐仅占总流量的1-3%，其中约85%被标记为欺诈并取消分成。Deezer CEO表示："没有其他公司跟随我们的步伐，所以我们决定让所有人都能检查自己的歌单。" > 💬 每天7.5万首AI音乐上传，人类创作的音乐正在被机器洪流淹没
来源：TechCrunch | 2026-06-11
Forrester警告：OpenAI可能成为"AI界的黑莓"
分析机构Forrester在OpenAI提交IPO申请之际发布报告，警告企业客户不要与单一AI供应商签订长期合同。报告称"OpenAI可能成为AI界的黑莓——定义品类的公司往往被品类革命最痛苦地淘汰"。Forrester建议企业"锚定你需要的能力，而非最早到达的品牌，保持低切换成本"。这番警告发布时，OpenAI正面临三重挑战：说服消费者使用其Agent、说服企业围绕其技术构建、在AGI竞赛中保持领先。 > 💬 黑莓当年也是"智能手机的王者"——然后iPhone来了
来源：The Register | 2026-06-11
微软总裁称Gen Z的AI反弹是"科技行业的警钟"
微软总裁Brad Smith发表博客文章，将大学毕业典礼上学生集体对AI发出嘘声称为"对科技行业的有力警醒"。Smith写道："对于那些追求用计算机替代人类的技术从业者，下一代给出了令人信服的回应：'没那么快'。"然而同一周，微软CFO向投资者透露员工人数同比下降且"预计趋势将继续"，同时微软计划在2026年投入约800亿美元AI基础设施。批评者指出这篇3000字的文章没有提出任何具体政策变化。 > 💬 左手写文章同情毕业生，右手砍人头投800亿搞AI——微软这波属于又当又立的教科书级示范
来源：Business Insider | 2026-06-11
甲骨文AI数据中心豪赌引发投资者恐慌，股价暴跌12%
甲骨文公布Q4营收192亿美元（同比增21%），但股价暴跌约12%。原因是资本支出飙升：2026财年capex达557亿美元，较上年的212亿美元暴涨163%。CFO指引2027财年净现金支出约700亿美元，并计划融资400亿美元（含200亿股权发行）。自由现金流赤字从2025年的3.94亿美元扩大到237亿美元。甲骨文押注与OpenAI和Meta的大规模数据中心交易来对抗AWS和Azure，但投资者显然对回报前景感到不安。 > 💬 一年花掉557亿美元建数据中心，自由现金流赤字237亿——甲骨文这是在用命赌AI
来源：The Register / Reuters | 2026-06-11
特斯拉前首席黑客创办AI网络安全公司Pi，估值1亿美元
特斯拉前内部黑客团队负责人Yoni Ramon创办网络安全初创公司Pi，估值1亿美元。Ramon曾为特斯拉服务六年，负责入侵车辆、机器人和太阳能产品，还在马斯克收购X期间保障其数据安全，被称为"马斯克十年来最爱的网络安全人"。Pi的AI系统充当"超级安全员工"，学习所有历史安全事件、公司政策和代码，实时监控内部开发者的工作以标记安全问题。xAI是Pi的首批客户之一，投资者包括CrowdStrike CEO和Armis联合创始人。 > 💬 从黑特斯拉到保护xAI——马斯克的御用黑客这次带着AI Agent出道了
来源：Forbes | 2026-06-10
DoorDash推出AI聊天机器人"Ask DoorDash"，拍照即可点外卖
DoorDash发布"Ask DoorDash"AI聊天机器人，支持用文字描述、照片或菜谱链接下单。用户可以说"一家四口的丰盛晚餐"或上传购物清单照片，机器人会自动构建购物车，还会提醒检查家里是否已有糖和黄油等常用食材。餐厅搜索也支持自然语言，如"适合孩子的素食餐厅"。iOS版已在部分地区上线，未来数周扩展至更多美国用户。 > 💬 终于不用在深夜饿着肚子翻菜单了——对着冰箱拍张照就能下单，外卖App的终极形态
来源：TechCrunch | 2026-06-11
Opendoor关闭印度业务引发AI取代外包大讨论
房地产科技公司Opendoor关闭其在印度钦奈和班加罗尔的办公室，CEO称要将运营工作迁回美国，转向"更小的AI原生团队"。印度目前拥有2100多个全球能力中心，雇佣约236万人，年收入近1000亿美元。风投Sheel Mohnot评论："随着手动工作被AI替代，印度将失去大量工作岗位。"分析人士认为，虽然Opendoor自身困境使情况复杂化，但"AI原生团队"的叙事标志着一个更广泛的转变正在发生。 > 💬 "AI原生团队"替代印度外包中心——1000亿美元的外包产业第一次感到了真正的威胁
来源：TechCrunch | 2026-06-11