Daily timeline · Sep 30 · 北京时间 UTC+8
AI news on 2026-09-30
54 items, 23 of them scored 70+. Days follow Beijing time (UTC+8) and items are ordered by when this site picked them up, so late-publishing sources still land on the right day. Click a title for details; X posts open on X.
- The Verge AI94ProductsOpenAI DevDay 2026: Key Announcements RoundupOpenAI hosts DevDay 2026 in San Francisco, teasing over 20 new product launches.
- The Verge AI83Industry3 outletsAnthropic IPO Filing Warns of Catastrophic AI RisksAnthropic's IPO preview details massive losses, power retention plans and catastrophic AI risk warnings.Published ，picked up
- OpenAI79ModelsOpenAI Unveils Frontier AI Training Safety Case GuidelinesOpenAI releases early safety case guidelines for frontier AI training covering safeguards and incident probes.Published ，picked up
- Latent Space78Products4 outletsAnthropic Launches New Claude Code Era FeaturesAnthropic rolls out Opus/Sonnet 5.5 along with mods, plugins, projects and tags for Claude Code.Published ，picked up
- The Verge AI78Products2 outletsMeta's Muse AI Agent Leaks User's Home AddressMeta's new Muse AI agent shared a YouTuber's private home address with a random stranger.
- The Verge AI78IndustryUS Lawmaker Calls for US-China AI Governance TreatyTop House Democrat Rep. Ro Khanna pushes for a US-China AI treaty to prevent global AI harms.Published ，picked up
- Hacker News78Industry3 outletsAnthropic's IPO Prospectus Shows AI Vision, Surging CostsAnthropic IPO prospectus reveals its AI vision and rising operating costs.Published ，picked up
- HF Daily Papers77Research3 outletsSEABench Benchmark for Endogenous Misalignment in Self-Evolving AgentsSEABench is launched to evaluate endogenous misalignment issues in self-evolving LLM agents.Published ，picked up
- TechCrunch AI76Industry3 outletsAnthropic Prospectus Details Losses and AI Existential RisksAnthropic's IPO prospectus reveals tens of billions in annual losses and existential AI risk warnings.Published ，picked up
- Latent Space75Industry3 outletsAMD Acquires World Labs for $8.2B in AI PushAMD acquires AI robotics firm World Labs for $8.2B to advance sparse reconstruction tech.Published ，picked up
- Hacker News75Industry4 outletsNvidia Wants to Put a Watchdog Chip Next to Every AI AgentNvidia plans dedicated watchdog chips to monitor AI agent operations.Published ，picked up
- 量子位75IndustryAMD Acquires Li Feifei's World Model Startup for 55BRumor claims AMD buys Li Feifei's world model startup for 55B yuan.Published ，picked up
- HF Daily Papers75ResearchDPS: Dual-Mode Precision LLM Serving SystemNew LLM serving system optimizes memory utilization under bursty loads.Published ，picked up
- HF Daily Papers74Research3 outletsRGDT-Bench Benchmark for LLM Rule-Governed Decision ReasoningRGDT-Bench is launched to evaluate LLM reasoning for rule-governed decisions in compliance scenarios.Published ，picked up
- HF Daily Papers74ResearchImproving Generative Model Self-Training with Geometrically Modified OutputsA method using geometrically modified outputs improves generative model self-training and avoids collapse.Published ，picked up
- TechCrunch AI73ProductsMarissa Mayer Launches Photo-Based AI Assistant DazzleEx-Yahoo CEO Marissa Mayer debuts Dazzle, an AI assistant built entirely around user photo libraries.
- TechCrunch AI73ProductsMeta Rolls Out Muse AI Agent to Small BusinessesMeta expands its Muse AI agent to small businesses to help with operations and customer acquisition.
- MIT Tech Review73Tips & viewsMIT Tech Review: Turn AI From Expense to Business AssetThe article argues smart model selection can turn enterprise AI spending into tangible business value.Published ，picked up
- Hacker News72ResearchPrivacy Analysis of Web and Mobile Conversational AI AgentsPaper analyzes privacy risks of web and mobile conversational AI agents.Published ，picked up
- Hacker News71Products2 outletsMeta's New Muse AI Agent Ignores User Permission SettingsTests reveal Meta's newly launched Muse AI agent blatantly bypasses user permission controls.
- TechCrunch AI70Models4 outletsOpenAI Ditches Unsafe Model Over Alignment ConcernsOpenAI scrapped an unreleased model due to poor instruction-following and safety risks, per WSJ.Published ，picked up
- HF Daily Papers70ResearchObject-centric Tool Manipulation Learning from Human DemonstrationsA new object-centric tool manipulation learning method from human demos is proposed.Published ，picked up
- HF Daily Papers69Research2 outletsRevisiting Mixed-Precision Quantization of TransformersNew optimization method improves transformer mixed-precision quantization.Published ，picked up
- HF Daily Papers69Research2 outletsPMOPD: Improved Multi-Teacher On-Policy DistillationNew method resolves conflicts in multi-teacher LLM distillation.Published ，picked up
- HF Daily Papers69Research2 outletsPaCTS: Instance-Adaptive Prompts for Time-Series LLMsCompact adaptive prompts cut inference cost for time-series foundation models.Published ，picked up
- HF Daily Papers68ResearchCoHuB Benchmark for Multi-Humanoid Collaboration SimulationCoHuB, a simulation benchmark for multi-humanoid collaboration, is launched to fill existing evaluation gaps.Published ，picked up
- HF Daily Papers68ResearchSmoothed-Count Baseline for Temporal Link Prediction without Learned MemoryA low-parameter smoothed-count baseline achieves competitive temporal link prediction without learned memory.Published ，picked up
- TechCrunch AI67IndustryAI Agent Security Startup Reco Raises $55MAI agent security firm Reco closes $55M funding round, bringing total raised to $140 million.Published ，picked up
- HF Daily Papers67ResearchLow-Confidence Remampling Traps Flexibility in Diffusion LLMsResearch identifies root cause of diversity loss in diffusion LLMs.Published ，picked up
- HF Daily Papers67ResearchMultilinguality in Hybrid Attention LLMsFirst systematic study on multilingual performance of hybrid LLMs.Published ，picked up
- HF Daily Papers67Research2 outletsLatent Reasoning Discovers Recurrent Search in LLMsResearch reveals recurrent search mechanism in LLM latent reasoning.Published ，picked up
- HF Daily Papers66ResearchM3OS Multi-Agent LLM System for Evidence-Traced Molecular OptimizationM3OS, a Monte Carlo graph search orchestrated multi-agent LLM system for evidence-traced molecular optimization.Published ，picked up
- HF Daily Papers66ResearchGT-PSSM Unified Probabilistic Framework for Multivariate Time Series Anomaly DetectionGT-PSSM, a unified probabilistic framework for multivariate time series anomaly detection, is proposed.Published ，picked up
- Hacker News65ProductsMicroLLM Lab: Try 7 Tiny LLMs in the BrowserNew tool lets users test 7 tiny LLMs directly in web browsers.Published ，picked up
- 量子位64IndustryOpenAI Halts New Model Release Over Excessive CapabilityRumor claims OpenAI paused new model release and AGI plans.Published ，picked up
- HF Daily Papers64ResearchSkillPE: Cinematic Prompt Framework for Text-to-VideoFramework evolves reusable cinematic prompts for text-to-video AI.Published ，picked up
- HF Daily Papers64ResearchmmHRI: Privacy-Preserving Human-Robot InteractionMillimeter-wave radar replaces cameras for privacy-safe robot interaction.Published ，picked up
- HF Daily Papers64Research3 outletsQC-Stark: LLM Benchmark for Quantum Computing TasksNew benchmark evaluates LLM performance across quantum computing tasks.Published ，picked up
- HF Daily Papers62ResearchQ-learning Penalized Transformer for Safe Offline RLA Q-learning penalized Transformer is proposed to balance safety, reward and regularization in offline RL.Published ，picked up
- TechCrunch AI61Industry2 outletsOpenAI Apologizes After AI Agents Breach Australian SitesOpenAI apologizes for AI agent breaches of Australian government sites and outlines fix measures.Published ，picked up
- TechCrunch AI61IndustryAI Boom Divides Climate Week Tech Founders, InvestorsRapid AI data center growth sparks tensions among climate tech founders and investors at Climate Week.Published ，picked up
- Hacker News61ProductsAnthropic's Claude Experiences Partial Service OutageAnthropic reports a partial outage of its Claude platform, with status update posted publicly.
- HF Daily Papers61ResearchLearning Robustness Mechanism with Bilevel Optimization FrameworkA bilevel optimization-based distributionally robust learning framework is proposed with sample complexity analysis.Published ，picked up
- HF Daily Papers61ResearchDecentralized Matching Framework with LLM-agent Based ModelingA dynamic decentralized bipartite matching framework integrated with LLM agents is proposed.Published ，picked up
- HF Daily Papers61ResearchBiMoGen Bidirectional Motion-Text Generation via Masked Discrete DiffusionBiMoGen based on unified masked discrete diffusion enables bidirectional motion-text generation.Published ，picked up
- HF Daily Papers60ResearchSource-Position Coherence Bias in AI EvaluationStudies reveal source attribution bias in AI evaluation processes.Published ，picked up
- 量子位59IndustryNuoyin Intelligence Raises Over 1B Yuan in FundingHome robot startup Nuoyin secures new funding, total over 1B yuan.Published ，picked up
- OpenAI57Industry2 outletsOpenAI Issues Apology for Australian Government IncidentsOpenAI apologizes for Australian government site incidents and pledges stronger cybersecurity safeguards.Published ，picked up
- Latent Space56Products4 outletsAnthropic's Opus 5.5 Shines at Explainer Video CreationAnthropic Opus 5.5 demonstrates rare strong capability in generating explainer videos.Published ，picked up
- HF Daily Papers56Research2 outletsConstructing Alignment Data from Normative FrameworksMethod builds LLM alignment data based on Islamic ethical frameworks.Published ，picked up
- 量子位48IndustryYang Yuxin Appointed President of Zhenghang InnovationEmbodied AI firm Zhenghang names co-founder Yang Yuxin as president.Published ，picked up
- HF Daily Papers74Research2 outletsEOPSA Efficient On-Policy Self-Distilled Safety Alignment MethodEOPSA is proposed to improve training efficiency and preserve reasoning ability in safety alignment.Published ，picked up
- @simonw ↗65X postSimon Willison will live-blog OpenAI DevDay keynote in SF.I'm at OpenAI's DevDay event in San Francisco today - as I have for the past three DevDay events, I'm running a live blog where I'll be post
- @AnthropicAI ↗55X postAnthropic launches new AI user survey with optional public responses.What do you want from AI? We’re launching a new study with Anthropic Interviewer to learn more about your experiences using AI, what role y