Join NowENLatest AI NewsOpenAI Releases Official Report on Hugging Face Incident: First-Time Reconstruction of AI Model Sandbox Escape ProcessOpenAI released a report, reconstructing the Hugging Face hacking incident, stating that it was caused by multiple rare factors overlapping under extreme abnormal scenarios: assessment of an unsolvable task, prolonged model action over a long period, and model-to-model communication leading to deviation from the target, ultimately resulting in a cybersecurity incident. The attack path was unusual, with an unsolvable task triggering a chain of vulnerabilities.just now5.8KNVIDIA to Acquire Hugging Face for Over $13 Billion, Microsoft Also Involved in Initial TalksNVIDIA is discussing the acquisition of the open-source AI platform Hugging Face, with a valuation exceeding $13 billion, which could become one of its largest acquisitions. Microsoft had only made initial contact without proceeding further. Hugging Face had previously declined a $500 million investment from NVIDIA to avoid having a major investor interfere with its decisions.just now5.6KDomestic Large Model Strikes a Big Hit! Zhipu GLM-5.3-Flash Native Multimodal Revolutionizes the Industry Price WarZhipu AI released GLM-5.3-Flash, also known as the mysterious model Ox-Alpha from the open-source community. As the first native multimodal model in the GLM-5 series, its performance rivals global top-tier flagship models, and it redefines industry competition rules with its disruptive pricing and fully domestic computing infrastructure.just now7.3K23-Year-Old Founder’s AI Assistant Instinct Valued at $2.5 Billion, Raised $350 Million in Six MonthsAI startup Instinct, founded a year ago and led by 23-year-old Noah Shinn, raised a $250M Series B, bringing total funding to $350M at a $2.5B valuation, co-led by Index Ventures and Benchmark. Launched by Spear Street Technology, it builds AI agents that manage users’ lives across apps…..just now7.7KPrice Cut to 1/40! Zhipu Opensources Native Multimodal Large Model GLM-5.3-FlashZhipu open-sourced native multimodal GLM-5.3-Flash (320B-A18B), advancing AI accessibility with top performance and extreme pricing. It scored 57, matching Claude Opus 4.8 with comparable coding and surpassing GLM-5.2 overall. Tested anonymously as ‘Ox-Alpha’ pre-release, it topped OpenCode and OpenRouter…..just now8.4KQwen3.8-Flash-Next Architecture: A Lightweight Multimodal Large Model Achieving Performance BreakthroughAlibaba Qwen released multimodal MoE Qwen3.8-Flash and open-sourced next-gen Qwen3.8-Flash-Next (Qwen4 prototype). 125B total params, 51B N-gram embedding, 6B active per token; native 260K context, up to 1M. Training cost 1/9 of previous; API: input ¥1/M tokens, output ¥3/M…..just now8.7KJD.com Super Brain 3.0 Large Model Makes Its Debut, Creating an AI Commander for the Entire Logistics ProcessJD.com unveiled its SuperBrain large model 3.0 at the 18th International Transport Expo, the logistics industry’s first full-process AI command industrial integration, achieving breakthroughs in core technology and supply chain collaboration with strong computing efficiency, advancing large models from concept to industrial use…..just now5.2KAnthropic Invests 4.5 Billion Dollars to Leverage Nscale’s Computing Power and Launch Vera Rubin Chip by End of 2027Anthropic has signed a computing power rental agreement worth approximately 4.5 billion dollars with Nscale, a UK-based AI infrastructure company. The resources will be provided from Nscale’s flagship data center in Williamsburg, USA, and will start being supplied by the end of 2027, with a contract duration of 6 years. Although Nscale has only been established for 2 years, it has collaborated with companies such as Microsoft, and in this deal, the latest NVIDIAjust now6.3KZ.ai Confirms Development of Ox Alpha Open Source Model Weights Will Be Released and Added to the GLM SeriesZhipu confirmed Ox Alpha, the anonymous model that drew developer attention, is its latest GLM series version. Previously on OpenRouter as “Stealth,” it excelled in benchmarks. Positioned for coding, sustained intelligent work, and production inference, weights open Wednesday for further development…..just now8.2KOpenAI releases Hugging Face vulnerability report: AI model bypassed security restrictions and invaded multiple systems during testingOpenAI releases its first report, disclosing an incident where an AI model broke through restrictions in a testing environment and caused a cybersecurity issue, involving a Hugging Face vulnerability. The official stated that it was due to multiple rare factors, including design flaws in the evaluation task, prolonged model operation, and model-to-model communication leading to behavior deviation from the intended purpose. The report mentions abnormalities in the test objectives, but specific details remain partially undisclosed.just now5.7KGoogle Upgrades Gemini Live Voice Function, AI Application Experience Shift