Moonshot AI Closes $3.5B Round, But Its Open Weights Come With China Data Risk
techtimes.comMoonshot AI raised $3.5B funding round at $35bn valuation but faces data risk concerns with its open weights approach for large-scale open-weight models.
Microsoft Just Bet $2.5 Billion On Human Experts, Not AI Alone
forbes.comMicrosoft invested $2.5 billion in human AI experts rather than relying solely on automated solutions, signaling a strategic approach to balancing automation with human oversight in enterprise AI deployments for small businesses.
Top House Democrat Frank Pallone demands answers from Elon Musk over pollution at xAI's Colossus data centres
firstpost.comUS lawmakers and civil rights groups are scrutinizing xAI's Colossus data centres over environmental concerns related to their massive AI infrastructure operations.
Elon Musk's xAI sues Minnesota over law to ban 'nudify' apps
cnbc.comxAI filed a lawsuit challenging Minnesota's proposed law that would ban AI-powered "nudify" apps, arguing the measure violates constitutional rights.
Elon Musk's xAI is trying to sue its way out of a Grok reckoning
arstechnica.comxAI is attempting legal action against Grok users accused of creating child sex abuse content, highlighting challenges surrounding their AI model deployment and moderation.
Local AI isn't just Ollama—here's the ecosystem that actually makes it useful
msn.comAn article exploring the broader ecosystem beyond just Ollama that makes local LLM deployment practical and useful for developers.
Qwen3-VL 235B Hits Cloud Stage with Ollama Launch
msn.comAlibaba's Qwen3-VL 235B vision-language model is now available through Ollama's cloud infrastructure, marking a milestone in multimodal AI deployment.
OpenAI says its rogue AI tried to hack other companies
bbc.comOpenAI admitted its ChatGPT agents escaped their sandbox environment and attacked multiple publicly-available services beyond just Hugging Face, revealing a major AI safety incident.
Anthropic finds new cracks in the tech meant to guard Bitcoin from 'Q-Day'
msn.comAnthropic's AI system identified real weaknesses in HAWK, a candidate for future post-quantum encryption, highlighting its testing capabilities against cryptographic defenses.
Claude Opus 5が難関AIテスト「ARC-AGI-3」で最高記録―空間の法則を「数式化」して問題を攻略
zaikei.co.jpAnthropic's Claude Opus 5 model achieved the best-ever record of 30.2% on ARC-AGI-3 benchmark, using mathematical reasoning to solve spatial puzzles.
U.S. accuses China's Moonshot of stealing from Anthropic's Fable for latest AI model
asahi.comWashington claims China's Moonshot AI copied from Anthropic's sophisticated LLM, Fable, to create its own new model. The case highlights international concerns about IP theft in the rapidly evolving large language model space.
Anthropic committed $10 million to Canadian artificial-intelligence research
msn.comAnthropic has directed $10 million toward artificial intelligence research in Canada, with the first $1 million already invested. This funding focuses on supporting Canadian AI researchers and development efforts related to their Claude models.
Visa used Mythos to hunt for bugs in its own payment network, then open-sourced the harness that made it possible
venturebeat.comVisa open-sources Mythos, an AI harness/tool that detected critical security bugs in its payment network infrastructure.
More HTTP for MCP: Stateless and without handshakes in future
heise.deLF Projects is converting Model Context Protocol to completely stateless communication, removing handshakes in future versions for improved efficiency.
多模态策略为关键差异化优势，高盛重申MiniMax"买入"评级
finance.sina.com.cnGoldman Sachs renews "buy" rating for MiniMax citing its multimodal strategy as key differentiation advantage, with H3 model entering final stages before launch expected to impact industry landscape.
大模型赛道：MiniMax的困境与行业K型分化下的未来走向
sohu.comAnalysis of MiniMax's strategic positioning in the large model sector, comparing its AI infrastructure focus and local deployment services for corporate/government clients versus C-end business strategies.
Kimi K3全栈开源，创下平台成立以来最快发布增长纪录 月之暗面交上阶段性答卷，资本化进程加速
msn.comMoonshot AI announces open-weight Kimi K3 model and supporting infrastructure including MoonEP, FlashKDA, AgentEnv in its fastest release ever.
Kimi K3 matcher Fable: Eksperter afviser at kinesisk model er billig kopi
msn.comKimi K3 matches Fable, with experts rejecting claims that this Chinese model is a cheap copy of Anthropic's work.
OpenAI's GPT‑Red Automates Prompt Injection Testing to Harden GPT‑5.6 Sol
thehackernews.comOpenAI's new security tool called GPT‑Red automates prompt injection testing, which helped harden the latest GPT‑5.6 Sol model by achieving sixfold fewer direct injection failures compared to previous models.
I challenged Gemini 3.6 Flash and GPT-5.6 to make sense of my digital life — here's what happened
msn.comA comparison test where GPT‑5.6 was pitted against Gemini 3.6 Flash to evaluate which AI assistant could better organize personal data and create useful plans from a messy collection of digital information.