Anthropic Escalates Cloud Spending With $45 Billion Nscale Agreement |...
pymnts.comAnthropic's $45 billion cloud computing lease agreement with Nscale escalates its infrastructure investment for AI model development and deployment.
C O R R E C T I O N -- OPAQUE industry collaboration on AI runtime verification standard
pr.cullmantimes.comAI industry leaders OPAQUE, AMD, Intel, and TII collaborate on TRACE, an open standard to advance runtime verification for AI systems.
Ocean City adopts AI policy amid legal, privacy concerns
yahoo.comA town has adopted its first local government policy governing artificial intelligence use and safety, setting guardrails around AI deployment in public sectors amid legal and privacy concerns.
Australia bans music made entirely by artificial intelligence from charts, awards show
kxan.comAustralia bans music made entirely by AI from ranking on official charts and being a contender at awards shows, marking regulatory action against generative content.
The Varied Perceptions and Experiences of Internal Versus External Auditors in Adopting Artificial Intelligence
cpajournal.comA study examining how internal and external auditors are adopting AI tools, revealing varied perceptions and experiences across the auditing industry.
Quollio, MCP server function officially released
sankei.comMCP server feature enables real-time collaboration between MCP-compatible LLMs/AI agents and metadata registered on QDIC as a bridge. AI can make tool calls to the MCP server hosted by Quollio.
Accelerated Understanding Inc. launches physics AI without relying on the Transformer model for research breakthroughs
msn.comCaltech startup launched a physics AI using neural operators instead of transformers, claiming 5 trillion evaluations in training time compared to traditional approaches. However, there is no benchmark proof yet that this alternative architecture outperforms transformer-based models on standard tasks.
GPUStack Day 0 supports Kimi-K3: vLLM and SGLang inference real test on 8×B300
163.comGPUStack Day 0 supports Kimi-K3 with performance comparison between vLLM and SGLang inference on an 8×B300 hardware setup, providing real-world benchmark results for open-source LLM serving.
Mystery Ox Alpha AI Model Sparks Zhipu, DeepSeek and Microsoft MAI Speculation
onmsft.comAn unidentified AI service with unusually generous rate limits appears on major API platforms including OpenRouter and OpenCode, sparking industry speculation about its identity among leading tech companies like Zhipu DeepSeek and Microsoft MAI teams.
智谱认领“牛来”并宣布上线、开源GLM-5.3
finance.sina.com.cnZhipu confirmed ownership of anonymous OxAlpha model that previously ran tests alongside DeepSeek and Microsoft MAI systems then launched open-source GLM-5.3. Prior platform trials generated high call volumes through OpenCode/Router services.
智谱：GLM-5.3-Flash跑在国产芯片上
finance.ifeng.comZhipu tested anonymous Ox-AI model on OpenCode platform with domestic chips powering calls before GLM-Flash launch. The model became most popular weekly offering both platforms highest demand while all request traffic used Chinese chip clusters for inference.
DeepSeek涨价，OpenCode扛不住了，用户怒喊退订
sohu.comDue to DeepSeek model price increases, OpenCode adjusted its usage quotas per five hours prompting user complaints about service cancellation. Users noticed reduced call limits after pricing changes affected their DeepSeek V4 series access through the platform.
OpenAI announces the Admin plugin for ChatGPT Work and Codex
9to5mac.comOpenAI announces new Admin plugin for ChatGPT Work and Codex, letting workspace admins manage users, analyze usage, and build reports.
Tragic mistake... Anthropic leaks Claude’s source code
msn.comAnthropic accidentally exposed source code for Claude Code, revealing vulnerabilities in its developer tool sandbox and triggering scrutiny over AI security practices. The incident highlights the risks of leaking sensitive infrastructure that developers rely on safely building with AI assistants.
Bloomberg: Alibaba Releases Smaller, Cost-Effective Qwen AI Model
bloomberg.comAlibaba releases smaller, cost-effective Qwen3.8 Flash model for global adoption with competitive pricing and open-weight features to challenge rivals like Meta's Llama models in the AI market.
China's AI Giants Embrace Open Weights, Putting Pressure on Closed-Source Rivals
yicaiglobal.comChina's major AI companies including DeepSeek and Alibaba release open-weight models, challenging closed-source alternatives like GPT-5.6 and Anthropic with more accessible model versions.
Google, Microsoft and Amazon Could Open the Door to China's Hottest AI Model
finance.yahoo.comMoonshot AI seeking partnerships with Google Cloud, Microsoft Azure and Amazon Web Services to host its Kimi K3 model for revenue generation.
Is Alibaba’s New $6 Spreadsheet-To-Video Tool Every Employee’s Worst Nightmare?
techround.co.ukAlibaba's Wan3.0 generates 1080p videos directly from documents for $6 per clip, representing a new approach to spreadsheet-to-video AI generation in the competitive text/video model space.
OpenAI Brings Back 5-Hour Codex Limit for ChatGPT Plus
techrepublic.comTechRepublic reports that OpenAI has restored a five-hour usage window limit on the Codex feature for ChatGPT Plus users, affecting developer productivity and tooling. Published on 2026-08-21.
Gemini's one billion users can't hide Google's cash flow crisis
finance.yahoo.comGoogle expands Gemini Enterprise AI platform for law firms, lawyers with enterprise-grade capabilities.