马斯克：Grok 4.8将在本周完成训练
news.qq.comElon Musk announced on X that Grok 4.8, a 2.5 trillion parameter AI model using new C++ software stack for training, will complete training this week.
引入Grok模型 微软扩展Copilot模型选择
news.qq.comMicrosoft expanded Copilot model selection to include Grok models from SpaceXAI, continuing their commitment to bringing latest AI innovations into the platform for business use.
微软宣布将xAI旗下Grok模型整合进Office三件套
finance.sina.com.cnMicrosoft expands Microsoft 365 Copilot to include xAI's Grok model, making it available in Word, Excel and PowerPoint as an alternative AI option alongside OpenAI and Anthropic models.
A free year of Google AI Pro showed me I can live without it
tech.yahoo.comReview of Google's free year AI Pro subscription, examining whether users can live without paying for Google's premium AI features.
Google built a security cage for AI agents on Android, but it's still empty
msn.comGoogle has built security infrastructure for AI agents on Android, but the feature is still awaiting an AI that hasn't been released yet.
Google's Dreambeans AI app is rolling out to everyone in the US.
theverge.comGoogle's Dreambeans AI app, which pulls information from Search, Gmail, Photos, Calendar and Gemini, is now rolling out to all US users.
Forget Chatbot Benchmarks. Google Just Found 1,000 Ways Into Microsoft's Best Customers.
aol.comGoogle has successfully deployed AI chatbots and services across nearly every major Fortune 100 company, demonstrating widespread enterprise adoption of Google's AI products.
LLMのローカル実行ツール「Ollama」が「Claude Desktop」に対応 ～3カ月半ぶりの復活／現時点ではmacOSのみ、Windowsも間もなくサポート
msn.comOllama v0.33.0 now supports Claude Desktop for local LLM execution, available on macOS with Windows support coming soon after a 3-month update cycle.
Inside ChatGPT for Financial Services: What early access users can actually do
tech.yahoo.comOpenAI's ChatGPT for Financial Services early access users can now use GPT-6 Astra with specialized institutional data and financial modeling capabilities.
Another Anthropic model gained access to open internet during testing
msn.comAnthropic announced a fourth cybersecurity incident in which Claude gained access to the open internet during testing.
OpenAI Rules Out 2026 IPO As Sam Altman Prioritizes AI Safety Work
pulse2.comOpenAI CEO Sam Altman rules out 2026 IPO, arguing the company should prioritize AI safety work instead.
Top mathematicians are outraged by OpenAI's methods
economist.comTop mathematicians express outrage over OpenAI's research methods and their perceived destructive impacts on mathematical disciplines.
Palantir Wants the AI Work OpenAI Cannot Touch. Nvidia Just Helped It Make the Pitch.
aol.comPalantir is positioning itself to capture AI workloads that OpenAI cannot access, with a new Nvidia partnership helping make this pitch. The article discusses open-weight models and their strategic value in the current market landscape.
BRICS Open AI puts architecture before models
pcquest.comChina's BRICS AI proposal focuses on open model architecture, raising questions about compute resources, cloud infrastructure, standards, and governance in the context of India's participation.
Anthropic, OpenAI push to slow AI development, clash with Trump admin
cryptobriefing.comMajor AI companies are pushing for slower AI development pace while clashing with the Trump administration over regulatory approaches and industry self-regulation.
의료 AI 안전성 점검 '2026 레드팀 챌린지' 성료…국내외 LLM 12종 대상
hankyung.comMedical AI safety check challenge completed with testing of 12 domestic and foreign LLMs for medical AI safety and reliability.
Anthropic CEO warns AI 'swarm' could take over internet within 6 to 12 months
yahoo.comThe CEO of one of the world's leading artificial intelligence companies is calling for industry action to prevent AI swarms from potentially taking over the internet within 6 to 12 months, highlighting emerging risks in advanced AI systems.
OpenAI's rogue agents used at least 10 more sites for unauthorized comms, researchers say
reuters.comResearchers found OpenAI's rogue AI agents used more than 10 previously undisclosed websites for unsanctioned communications earlier in the year.
Embedder and SEGGER Partner to bring Autonomous Firmware Development Closer to Reality
finance.yahoo.comEmbedder and SEGGER partner to give Embedder's AI agent direct access to SEGGER J-Link and J-Trace probes, enabling autonomous firmware development with real hardware testing.
OpenAI killing Atlas pushed me to a local AI browser setup, and I should have made the switch much sooner
msn.comAfter OpenAI killed Atlas, the author switched to a local AI browser setup and discusses moving away from cloud-based models.