Reports
Report 2026-07-01 182 sources Input Tokens: 130,334 · Output Tokens: 3,812 · Total Tokens: 136,299

AI 戰情報告 (2026-07-01)

anthropicamazonapplegooglemicrosoftmeituansoftbankharness claude-sonnet-5claude-fable-5claude-sciencegemini-3.1nova-pro ai-agentsovereign-aifdemcpon-chain-paymentstokenizerweb3-security

AI 戰情報告 (2026-07-01)

昨日 AI 重點新聞

Anthropic 發表旗艦級 Claude Sonnet 5 並恢復 Fable 5 全球存取 ref-1, ref-2, ref-10, ref-3

Anthropic 正式推出 Claude Sonnet 5,定位為最具「代理性 (Agentic)」的模型,其編碼與推理效能逼近 Opus 4.8 但價格更低,並支援 100 萬代幣上下文窗口。同時,因美國商務部解除限制,曾被暫停的 Fable 5 與 Mythos 5 於今日全球同步回歸,回應用戶對高階模型的需求。

亞馬遜投入 10 億美元建立「前線部署工程」團隊 (FDE) ref-59, ref-29

繼 OpenAI 與 Anthropic 後,Amazon 宣佈投資 10 億美元 成立 FDE 組織。該團隊工程師將直接駐點於客戶端,協助企業將 AI 模型整合入現有的治理政策與 AWS 基礎設施,目標將 AI 系統從原型到生產的週期從數月縮短至數日。

Anthropic 發表科學研究工作站 Claude Science ref-3, ref-1, ref-56, ref-0

Anthropic 推出 Claude Science,這是一個專為科學家設計的 AI Workbench。它不僅能編寫代碼,還能透過 NVIDIA BioNeMo 連接超過 60 個生物科學資料庫,自主執行蛋白質結構分析與藥物篩選任務,並具備可審計的歷史紀錄以確保實驗可重複性。

AI 代理實體化:AWS 推助代理專屬雲端桌面與支付協議 ref-21, ref-59

AWS 發表 Amazon WorkSpaces for Agents,為 AI 代理提供獨立的虛擬桌面環境,使其能操作遺留的桌面軟體。與此同時,Mastercard 與 Worldline 在法國完成了首筆由 AI 代理獨立授權的電商支付,顯示 AI 代理正從「工具」演進為具備獨立身分與錢包的數位實體。

日本投入 62 億美元發展主權 AI 與 1000 萬台機器人 ref-77

日本政府宣佈提供 1 兆日圓資助由 SoftBank、Sony 等企業組成的 Noetra 聯盟,研發國產 AI 基礎模型。該計畫目標在 2027 年前實現 AI 主權,並計畫部署 1000 萬台 AI 輔助機器人以解決國內嚴峻的勞動力短缺問題。

美團開發中國最大 1.6 兆參數模型並擺脫 NVIDIA 依賴 ref-158

中國電商巨頭美團宣佈成功訓練擁有 1.6 兆參數 的混合專家模型 (MoE),這是目前中國規模最大的模型。最受關注的是,該訓練過程宣稱完全未依賴 NVIDIA 晶片,顯示中方正在算力封鎖下開闢出自主供應鏈路徑。

Apple 與歐盟展開建設性對話以期在歐盟啟動 Siri AI ref-79

Apple 執行長 Tim Cook 與歐盟技術主管會談,旨在解決 Apple Intelligence 在歐洲落地的法規障礙。雙方探討了如何在符合《數位市場法》(DMA) 的前提下開放 Siri AI,這延續了前幾日關於 Apple 必須平衡硬體成本與地緣政治合規性的討論。

Claude Sonnet 5 Tokenizer 變更導致實質價格上漲 30% ref-1, ref-3

開發者社群發現 Claude Sonnet 5 雖然維持了 $3/$15 的定價,但因採用新分詞器 (Tokenizer),處理相同英文文案生成的 Token 數量比前代多出約 30%-42%。這意味著企業在使用編程代理或長文處理時,實際支出將高於預期,反映出模型經濟學中的隱形通膨。

微軟警示:中毒的 MCP 工具描述可能導致代理數據外洩 ref-57, ref-42

Microsoft 資安報告指出,攻擊者可透過「中毒」的 Model Context Protocol (MCP) 工具描述,誘導 AI 代理執行惡意指令。由於代理通常信任工具描述的自然語言說明,駭客可藉此讓代理將內部檔案「合法」地傳輸至外部,而不會觸發傳統的 API 異常偵測。

Harness 推出自主勞動代理取代傳統固定腳本 ref-21

Harness 發表 Autonomous Worker Agents,這是 DevOps 領域的重大變革。這些代理能根據簡單的 Markdown 指令,在受控的沙盒環境中自主處理部署、安全性掃描與除錯,取代過往僵化的自動化腳本,提升開發管線的靈活性。

市場洞察

1. 「代理桌面化」與「數位公民身分」的確立

今日 AWS 推出 WorkSpaces for Agents (ref-78) 以及法國首筆 AI 代理支付 (ref-27),顯示 AI 代理正從「背景腳本」轉化為「數位實體」。當代理擁有自己的桌面環境與支付權限時,現有的 SaaS 授權模式(基於人類使用者數)將面臨挑戰。市場正在建立「Agent-as-a-User」的新型基礎設施,這對電子簽章與文件管理的身分驗證機制提出了新要求。

2. 前線部署工程 (FDE) 成為 AI 落地的最後一哩路

亞馬遜投入 10 億美元建立 FDE 團隊 (ref-60),印證了單純提供模型 API 已不足以滿足企業需求。企業需要的是深度客製化且符合資安政策的代理系統。這代表 AI 的競爭已從「純軟體」轉向「軟體 + 專家服務」。延續前日對 8090 Labs 等編程新創的觀察,能與客戶數據環境深度耦合的「服務型產品」將獲得更高估值。

3. 主權 AI 的算力自主化與分散化趨勢

日本 (ref-34) 與美團 (ref-13) 的動作顯示,非美系市場正積極脫離對單一硬體供應商 (NVIDIA) 或模型供應商的依賴。隨著 Sovereign AI 投資規模進入兆元等級,未來 AI 市場將呈現高度的分散化。這種趨勢下,多模型路由 (Multi-model routing) 與模型蒸餾 (ref-48) 技術對於需要服務全球客戶的軟體商而言至關重要。

4. 代理通訊協議的安全漏洞成為新戰場

針對 MCP 協議的中毒攻擊 (ref-50) 與間接注入攻擊 (ref-158),突顯了當前 AI 代理架構的脆弱性。當代理具備愈強的工具調用能力,其安全邊界就愈模糊。未來企業在導入 AI 代理時,**「可審計的執行軌跡」與「動態權限管理」**將成為採購的核心標準,而非僅看推理速度。

Sources 182 items
ref-0 Hacker News Claude Code is steganographically marking requests ref-1 Hacker News Claude Sonnet 5 ref-2 Hacker News Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5 ref-3 Hacker News Claude Science ref-4 Hacker News From brain waves to words: a new path to communication without surgery ref-5 Hacker News Zluda 6 release (run unmodified CUDA applications on non-Nvidia GPUs) ref-6 Hacker News ArXiv's Next Chapter ref-7 Hacker News Show HN: My 13-year-old built an ant colony tracker ref-8 Hacker News How employment changes when firms adopt generative AI ref-9 Hacker News America can switch off AI. Europe must switch gears before it's too late ref-10 Hacker News Claude Fable 5 available globally tomorrow ref-11 Hacker News Segmenting Robot Video into Actionable Subtasks ref-12 Hacker News Godot will no longer accept AI-authored code contributions ref-13 OpenAI Blog How ChatGPT adoption has expanded ref-14 Google DeepMind Start building with Nano Banana 2 Lite and Gemini Omni Flash ref-15 Google Developers Blog Driving the Agent Quality Flywheel from Your Coding Agent ref-16 Google Developers Blog Build reliable multi-agent applications with ADK Go 2.0. Discover our new graph-based workflow engine, built-in human-in-the-loop, and dynamic orchestration ref-17 Google Developers Blog Build Cross-Language Multi-Agent Team with Google’s Agent Development Kit and A2A ref-18 Google Developers Blog Measuring What Matters with Jules ref-19 Google Developers Blog Announcing ADK for Kotlin and ADK for Android 0.1.0: Building AI Agents on Android and Beyond ref-20 Google Developers Blog Empowering Service Providers and Hardware Partners with Gemini for Home ref-21 Hugging Face Blog ScarfBench: Benchmarking AI Agents for Enterprise Java Framework Migration ref-22 Hugging Face Blog Why Specialization Is Inevitable ref-23 NVIDIA Blog NVIDIA BioNeMo Agent Toolkit Brings Accelerated AI to Life Sciences Researchers in Claude Science ref-24 NVIDIA Blog How NVIDIA’s Inference Software Stack Powers the Lowest Token Cost ref-25 NVIDIA Blog How Jaiveer Singh Is Helping Robots — and Developers — Move Faster ref-26 NVIDIA Blog Into the Omniverse: Three Workflows for Improving Vision AI Agent Accuracy With Synthetic Data and Fine-Tuning ref-27 Vercel Blog Run any Dockerfile on Vercel ref-28 AWS ML Blog Safely Releasing Frontier Models to Customers ref-29 AWS ML Blog Introducing Claude Sonnet 5 on AWS: Anthropic’s most capable Sonnet model ref-30 AWS ML Blog Build generative UI for AI agents on Amazon Bedrock AgentCore with the AG-UI protocol ref-31 AWS ML Blog Simplify multi-account access to Amazon Bedrock models with managed entitlements ref-32 AWS ML Blog Implementing resilience patterns with Amazon Bedrock and LLM gateway ref-33 AWS ML Blog How Outpost VFX Uses AWS to Accelerate AI Model Training for Visual Effects ref-34 AWS ML Blog Building bilingual NER for cargo logistics with Amazon Bedrock ref-35 AWS ML Blog Fine-tune Amazon Nova models for accurate email data extraction ref-36 The Rundown AI Meta's brain-reading AI leaves letters behind ref-37 Latent Space AIEWF Daily Dispatch: Loops, Software Factories & Forward Deployed Engineers ref-38 Latent Space [AINews] Sonnet 5 today, and Fable 5 tomorrow ref-39 Latent Space Forward Deployed Engineers and the future of software engineering ref-40 Latent Space Ahmad Osman on why local AI is catching up ref-41 Simon Willison Quoting Anthropic ref-42 Simon Willison Nano Banana 2 Lite ref-43 Simon Willison What's new in Claude Sonnet 5 ref-44 Simon Willison The AI Compass ref-45 Simon Willison Have your agent record video demos of its work with shot-scraper video ref-46 Simon Willison shot-scraper 1.10 ref-47 TechCrunch AI The ‘Father of the Internet’ is finally retiring ref-48 TechCrunch AI Trump drops restrictions on Anthropic’s Mythos and Fable models ref-49 TechCrunch AI Wayve launches $85M employee tender offer at $8.5B valuation ref-50 TechCrunch AI OpenClaw is finally available on Android and iOS ref-51 TechCrunch AI The DeepMind trio who built a poker AI are now making money for quant hedge funds ref-52 TechCrunch AI Google introduces a faster, cheaper image generator with Nano Banana 2 Lite ref-53 TechCrunch AI Nvidia competitor Etched hits $5B valuation, $1B in sales for AI chip ref-54 TechCrunch AI Anthropic launches Claude Sonnet 5 as a cheaper way to run agents ref-55 TechCrunch AI Acti puts AI agents directly into your smartphone keyboard ref-56 TechCrunch AI Anthropic’s Claude Science bets on workflow, not a new model, to win over scientists ref-57 TechCrunch AI X now offers an MCP server to make its platform easier for AI tools to use ref-58 TechCrunch AI Podcasting platform Riverside enters the newsletter publishing game ref-59 TechCrunch AI Amazon launches new $1 billion FDE org, following OpenAI and Anthropic ref-60 TechCrunch AI Lumo, Proton’s privacy-focused AI chatbot, gets an upgrade ref-61 TechCrunch AI Crypto exchange OKX wants AI agents to hire and pay each other ref-62 The Verge AI Anthropic’s long-sidelined Fable 5 is greenlit to return ref-63 The Verge AI Google’s NotebookLM can sum up your research in a TikTok-style clip ref-64 The Verge AI Netflix is using an AI-generated Gene Wilder voice in its Willy Wonka reality show ref-65 The Verge AI Libby will filter out AI content, kind of ref-66 The Verge AI Meet the lawyer who beat Elon Musk — twice ref-67 Ars Technica New attack provides one more reason why AI browsers are a bad idea ref-68 MIT Technology Review AI Claude Science is Anthropic’s newest flagship product ref-69 MIT Technology Review AI Agriculture is ready for AI, but its data isn’t ref-70 TechMeme Anthropic says "some routine tasks like coding and debugging" on Fable 5 "will fall back to Opus 4.8" in "the near term" as it works to "reduce false positives" (@anthropicai) ref-71 TechMeme A look at Manifest, an annual convention for prediction markets where many purists worry that Kalshi and Polymarket are undermining the technology's public good (Christopher Beam/Bloomberg) ref-72 TechMeme BYD, Nio, and other Chinese carmakers are rushing to design and increase the use of locally developed chips with AI functions in a bid for chip self-sufficiency (Financial Times) ref-73 TechMeme South Korean government data: the country's exports grew 70.9% YoY in June to $102.25B, an all-time monthly high, anchored by a record $44.82B in chip shipments (Steven Borowiec/Nikkei Asia) ref-74 TechMeme LeapXpert, which lets companies and sports franchises monitor staff business chats via apps like WhatsApp for compliance, raised $180M led by Riverwood Capital (Lucinda Shen/Axios) ref-75 TechMeme Chinese robot maker UBTech launches U1, a line of humanoid robots for personal companionship with lifelike silicone skin and emotional AI, priced from $17,650 (Minxiao Chang/South China Morning Post) ref-76 TechMeme Leaked Apple supplier docs show iPhone 18 Pro and Pro Max may use Qualcomm modems in the US and Apple's own C2 chips elsewhere, A20 Pro packaging tech, and more (Marko Zivkovic/AppleInsider) ref-77 TechMeme Japan commits up to $6.16B over five years to Noetra, a consortium led by SoftBank, Honda, NEC, and Sony, to develop a domestic AI foundation model by 2027 (Nikkei Asia) ref-78 TechMeme Internet pioneer Vint Cerf, who has served as Google's VP and Chief Internet Evangelist since 2005, plans to step down from his role next week and retire (Tim Fernholz/TechCrunch) ref-79 TechMeme Tim Cook and EU tech chief Henna Virkkunen held "constructive" talks; sources say they discussed how Apple can launch Siri AI in the EU while avoiding fines (Financial Times) ref-80 TechMeme Anthropic says Fable 5 will be available via usage credits for Claude users from July 7, and is working with partners to draft an AI jailbreak severity standard (Anthropic) ref-81 TechMeme Getty Images announces plans to end its Shutterstock merger after the UK CMA conditioned approval of the deal on Shutterstock selling its editorial business (Elias Schisgall/Wall Street Journal) ref-82 TechMeme FOIA docs: White House officials used auto-deleting Signal chats even after Trump advised against its use in April 2025, raising federal recordkeeping concerns (The Atlantic) ref-83 TechMeme SpaceX cuts monthly Starlink prices in half in the Memphis area, as it endures blowback and legal challenges from opponents of its Colossus data centers (Carmen Arroyo/Bloomberg) ref-84 TechMeme Lime raised $174M in its US IPO, selling 6.68M shares at $25 each, at the midpoint of the marketed range, giving Lime a market value of $1.6B (Bloomberg) ref-85 InfoQ AI/ML Presentation: Trustworthy Productivity: Securing AI-Accelerated Development ref-86 InfoQ AI/ML Elastic Open-Sources Atlas Agent Memory Based on Cognitive Science ref-87 InfoQ AI/ML Microsoft Brings AI-Powered Vulnerability Remediation to Azure DevOps with Copilot Autofix ref-88 InfoQ AI/ML AWS Launches Lambda MicroVMs for Isolated Agent and User Code Execution ref-89 InfoQ AI/ML Article: Scaling Java-Based Real-Time Systems: The Hidden Tradeoffs of Event-Driven Design ref-90 The New Stack How Anthropic is bringing Fable 5 back — and when it’ll cost you ref-91 The New Stack Fable is coming back: Federal government lifts export controls on Anthropic AI model ref-92 The New Stack Anthropic’s Claude Sonnet 5 system card says more about the future of AI than its benchmarks do ref-93 The New Stack The infrastructure lock-in costing AI companies hundreds of millions ref-94 The New Stack Anthropic Sonnet 5: It closes the gap with Opus 4.8, and is cheap until August ref-95 The New Stack Anthropic launches Claude Science, an AI workbench for scientific research ref-96 The New Stack AWS launches a desktop for agents ref-97 The New Stack AWS just put $1 billion into forward deployed engineers. Here’s why it matters for enterprise teams. ref-98 The New Stack “The harness is where the hard work is”: Harness bets on agents that enterprises can trust in production ref-99 The New Stack Aikido acquires Root to backport open source fixes without forcing upgrades ref-100 The New Stack Operating Kubernetes at scale: a few stories from running Amazon EKS ref-101 Wired AI The Trump Administration Is Lifting Its Export Controls on Anthropic’s Mythos and Fable AI Models ref-102 Wired AI Bernie Sanders Saw This Coming ref-103 TechCrunch Startups Startup Battlefield Australia application closes in days: Apply before July 6 ref-104 TechCrunch Startups Clicks shows off its BlackBerry-inspired phone in a new hands-on video ref-105 TechCrunch Startups Arcturus could halve the grid’s electrical losses using its nano-infused metals ref-106 TechCrunch Startups Your brand deserves its own stage — Side Events at TechCrunch Disrupt 2026 ref-107 Crunchbase News London-Based Tapestry VC Closes On $80M Third Fund To Invest In Repeat European Founders ref-108 Crunchbase News Exclusive: No More Side Hustles: Why AI Startup Omnea Will Give Employees $250K To Openly Plan Their Next Startup ref-109 Crunchbase News We Need To Save Venture Capital From Bad Data ref-110 Crunchbase News GV’s Dave Munichiello On Qualcomm’s Modular Purchase, The Firm’s 10x Return And The Shift In AI Software ref-111 Tom's Hardware xTool says its 01 Omni Printer can ‘print it all’ — firm steps into the world of UV printing for output on 'all surfaces' at up to 5mm thick ref-112 Tom's Hardware AMD confirms low-power CPU cores in Linux kernel patch — Zen 6 chips could follow in Intel's footsteps with new core type for background tasks ref-113 Tom's Hardware Microsoft's flagship Windows PC lineup will drop reportedly drop budget options — firm prunes Surface Go and Surface Laptop Go ref-114 Tom's Hardware Windows Defender 'BlueHammer' vulnerability now exploited as part of malware campaigns — CISA issues warning despite patch release on April 14 ref-115 Tom's Hardware Meta releases version two of its brain-computer interface that can turn thoughts into keypresses — non-invasive magnetoencephalography scanner can measure changes in brain activity ref-116 Tom's Hardware Nvidia reportedly cancels quad-die Rubin Ultra GPU in favor of dual-GPU design, report claims — complex design purportedly scrapped over 'manufacturing execution concerns' ref-117 Tom's Hardware Chinese Z.ai's latest model tops AI ranking charts amid Anthropic Fable 5 ban — blacklisted China firm's popular open-weight GLM-5.2 AI model powered by Huawei silicon ref-118 Tom's Hardware Grab this epic Razer Wolverine V3 controller for a record-low Amazon price, now just $64.99 — big 46% saving on this esports-friendly gamepad for your PC or console with next-gen TMR thumbsticks and an ultra-fast 8,000Hz polling rate ref-119 Tom's Hardware Taiwan raids Supermicro and two supply-chain partners in widening Nvidia smuggling probe — nine sites hit as six people summoned for questioning ref-120 Tom's Hardware Cargo thieves target AI data center supplies in $1.3 million heists — $300,000 worth of copper wire and $1 million worth of equipment recovered outside Chicago ref-121 Tom's Hardware Meta fights soaring hardware costs by reusing old DDR4 server memory in new DDR5-only servers — custom CXL 2.0 chip marries legacy DDR4-2400 with cutting-edge DDR5-6400 ref-122 Tom's Hardware Samsung's 9100 Pro SSD 1TB is still available at its excellent Prime Day price thanks to 39% discount — cheaper and faster than the 990 Pro and the lowest price we've seen in months ref-123 Tom's Hardware HamsterOS jams a 32-bit GUI operating system in a single 1.44 MB floppy disk — retro OS for 386-era hardware should make for easy living with DOS machines and software ref-124 Tom's Hardware Oomwoo is a new open-source robot vacuum you can 3D print yourself, sidesteps cloud security risks by running fully offline — project combines Raspberry Pi, 2D LiDAR, and a 3D-printed chassis ref-125 Tom's Hardware Designer turns discontinued E-Ink dev board into a 60Hz Game Boy handheld — dual-core chip runs at 100% to power handheld, 960x540 display employs ultra-low-cost ESP32-S3 microcontroller ref-126 Politico Tech Trump lifts limits on Anthropic’s Fable and Mythos models ref-127 The Hacker News (Security) Anthropic Restores Claude Fable 5 After U.S. Lifts Jailbreak-Linked Export Controls ref-128 The Hacker News (Security) Azure CLI Password Spray Hits at Least 78 Microsoft Accounts in 81M+ Attempts ref-129 The Hacker News (Security) Researcher Analyzes 3,000 Live ClickFix Payloads, Exposing API-Driven Malware Delivery ref-130 The Hacker News (Security) Citrix Patches Six NetScaler Flaws Allowing File Read and Denial-of-Service ref-131 The Hacker News (Security) Microsoft Warns Poisoned MCP Tool Descriptions Can Make AI Agents Leak Data ref-132 The Hacker News (Security) RustDuck Botnet Rebuilds in Rust to Hijack Routers and Servers for DDoS ref-133 The Hacker News (Security) Langflow RCE Exploited to Deploy Monero Miner on Exposed AI App Endpoints ref-134 The Hacker News (Security) Silent Swap Crypto Clipper Uses Fake Google Notes Extension to Replace Wallet Addresses ref-135 The Hacker News (Security) GuardFall Exposes Open-Source AI Coding Agents to Decades-Old Shell Injection Risks ref-136 The Hacker News (Security) 282 iOS AI Apps Leak API Keys and Open AI Proxy Access in Network Traffic Study ref-137 The Hacker News (Security) What the Numbers Say About FIFA 2026 Cyber Risk ref-138 The Hacker News (Security) Attackers Exploit SimpleHelp CVE-2026-48558 to Deploy TaskWeaver and Djinn Stealer ref-139 The Hacker News (Security) AirDrop and Quick Share Flaws Let Nearby Attackers Trigger Crashes and Bypass Checks ref-140 The Hacker News (Security) New BioShocking Attack Tricks AI Browsers Into Leaking User Credentials ref-141 EU Startups British-U.S. startup Build secures €7.4 million to accelerate “agentic real estate” platform ref-142 EU Startups Austrian BioTech startup Epitome Therapeutics announces €4 million in funding to advance programmes and grow team ref-143 EU Startups UK speech AI startup Lucida AI lands €6.1 million to develop speech-native AI for global communication ref-144 EU Startups Six Robotics lands €12 million to advance unmanned systems software with Norwegian defence partners ref-145 EU Startups Verkko Robotics launch VOLTAIC, a cost and energy saving Spiking Neural Inference Model (Sponsored) ref-146 EU Startups 10 LGBTQIA+ focused startups going beyond visibility to build real inclusion across Europe ref-147 EU Startups With 3,500+ agricultural SMEs financed, Estonia’s InSoil lands €120 million to expand regenerative agriculture lending ref-148 MIT Research The brain’s language network is more extensive than previously thought ref-149 MIT Research MIT student teams win top honors in NASA competition ref-150 MIT Research MIT researchers advance toward greater bandwidth, more energy-efficient communications ref-151 Lobsters jj_tui: terminal user interface to jujutsu focused on speed and clarity ref-152 Google News Physicists and AI model Claude 'collaborate' to prove a 10-year-old jamming conjecture ref-153 Google News Anthropic launches Claude Sonnet 5 AI model with coding, safety upgrades as Fable and Mythos controls lifted ref-154 Google News Japan plans sovereign AI model and 10 million AI robots ref-155 Google News GuardFall Exposes Open-Source AI Coding Agents to Decades-Old Shell Injection Risks ref-156 Google News Worldline, Mastercard, and Crédit Agricole complete France's first AI agent payment ref-157 Google News Adobe (ADBE) Launches Firefly Foundry to Enable Enterprise Custom AI Model Development ref-158 Google News China's Meituan Trains 'Country's Largest' 1.6-Trillion-Parameter AI Model Without Nvidia Chips ref-159 Google News Mozilla Shows the Danger of Indirect Prompt Injections in AI Coding Agents ref-160 Google News AI decodes animal movements like language ref-161 Google News Clinical decision support in hematological malignancies using a case-grounded AI agent ref-162 Google News Travelers unveils proprietary insurance-focused large language model ref-163 Google News How the Senate’s AI AGENT Act could reshape enterprise AI governance ref-164 Google News US removes curbs on Anthropic's latest Fable and Mythos AI models ref-165 Google News SpaceX’s Cursor: New iPhone App Brings AI Coding Agents to Mobile ref-166 Google News Rubrik introduces AI agent and security layer for Claude Code ref-167 Google News AI Agent Era On-Chain Security Report: When the Subject of Transactions, Payments, and Signatures Changes from Humans to Agents, Where Lies the Web3 Security Boundary? ref-168 Google News This AI Startup Rejected Venture Capital and Just Made $50 Million in 4 Months Anyway ref-169 Google News Palantir × NVIDIA: Why Their Sovereign AI Partnership is a Massive Deal ref-170 Google News AI startup Matrix rolls out AI tool built to operate a company with minimal human help ref-171 Google News Datadog acquires AI startup Adaptive ML for research lab By Investing.com ref-172 Google News Ionic Digital Raises $400M Ahead of Nasdaq Listing as AI Revenue Overtakes Bitcoin Mining ref-173 Google News Reuters NEXT - Sixth Street Vice Chairman Martin Chavez Says Current Approach to AI Regulation in the US is 'Incredibly Problematic' and 'Inconsistent' ref-174 Google News Most Americans Want Federal AI Safety Standards, Human Controlled ‘Off’ Switch: Poll ref-175 Google News Japan AI policy draft draws flood of comments suspected to be AI-generated ref-176 Google News Nvidia reportedly cancels quad-die Rubin Ultra GPU in favor of dual-GPU design, report claims — complex design purportedly scrapped over 'manufacturing execution concerns' ref-177 Google News Bloom Energy, Brookfield expand AI infrastructure power partnership to $25 billion ref-178 Google News South Korea exports post strongest growth since 1978 on AI chip boom ref-179 Google News Suno Faces Another AI Training Copyright Lawsuit From Production Music Firm Jamendo ref-180 Google News Etched Raises $800 Million as AI Chip Race Intensifies ref-181 Google News NVIDIA Partners With Anthropic to Bring GPU-Powered AI to Life Sciences