AI news of October 28, 2026

By field: Products & applications 20, Chips & compute 4, Policy & regulation 3, Funding & companies 2, AI × finance 1, AI for Science 1, AI × humanities & society 1

• 23:55GitHub Copilot App Launches Projects Feature

  GitHub Copilot app introduces project functionality to unify work across multiple repositories. Users define tasks, add context like design mocks or Slack threads, and select…

  github-changelog-copilotProducts & applicationsoriginal ↗

• 23:50GitHub Enterprise now includes Copilot

  GitHub Enterprise now includes Copilot access for all licensed users. AI usage is metered from first request against enterprise budget with no included credits. Code completion…

  github-changelog-copilotProducts & applicationsoriginal ↗

• 23:45GitHub Copilot enhances cross-client session continuity in VS Code

  GitHub enhances Copilot in VS Code with cross-client session continuity, allowing users to start and continue agent sessions across devices. Sessions remain active even when…

  github-changelog-copilotProducts & applicationsoriginal ↗

• 23:45GitHub Copilot launches security extension

  GitHub has launched a security extension for Copilot in public preview, enabling users to view and manage code scanning alerts directly within the Copilot app. Alerts are ordered…

  github-changelog-copilotProducts & applicationsoriginal ↗

• 23:41South Korea Faces New Cyberattacks on Financial Firms

  South Korea's financial institutions face a new wave of cyberattacks, affecting Daishin Securities, Mirae Asset Global Investments, and Bithumb. Researchers suspect a China-linked…

  TechCrunchAI × financeoriginal ↗

• 23:40GitHub Intelligence Enhances Development Context

  GitHub Intelligence uses its data graph to provide context for AI agents, enabling them to understand team workflows and code history. Copilot Memory now remembers developer…

  github-changelog-copilotProducts & applicationsoriginal ↗

• 23:30GitHub Copilot generates and edits images via natural language

  GitHub Copilot now generates and edits images via natural language using GPT Image 2.5 Flare. Images are previewed inline and saved directly to the workspace. The feature is…

  github-changelog-copilotProducts & applicationsoriginal ↗

• 23:30GitHub Copilot Fusion supports local model routing

  GitHub Copilot Fusion now enables intelligent routing between local and cloud models in experimental preview. Users can select on-device or cloud models via Copilot CLI, app, or…

  github-changelog-copilotProducts & applicationsoriginal ↗

• 23:30GitHub Copilot Launches New Features in JetBrains

  GitHub has enhanced its Copilot integration in JetBrains IDEs with Copilot Fusion, local sandboxing, and Next Edit Suggestions. Copilot Fusion automatically selects models based…

  github-changelog-copilotProducts & applicationsoriginal ↗

• 23:00RTX Spark Laptops Get New Models and Games

  NVIDIA RTX Spark laptops from ASUS, Dell, HP, Lenovo, Microsoft and MSI now feature new AI models and gaming enhancements. The systems integrate Blackwell RTX GPUs and Grace CPUs…

  NVIDIA NewsroomChips & computeoriginal ↗

• 22:58NVIDIA Releases Nemotron 3.5 Transcribe for Multilingual Transcription

  NVIDIA releases Nemotron 3.5 Transcribe, an open-weight multilingual transcription model supporting 40 languages, speaker diarization in 13, and translation across 31 languages.…

  Hugging Face BlogProducts & applicationsoriginal ↗

• 20:47MIT Smaragdis Develops AI That Outperforms Humans in Sound Analysis

  MIT's Paris Smaragdis research team develops AI that outperforms humans in sound analysis, focusing on separating and interpreting complex audio environments. The system can…

  Korea AI TimesPolicy & regulationoriginal ↗

• 20:01Amazon Launches New Echo Devices for Alexa Ecosystem

  Amazon launched new Echo devices including Echo Dot, Echo Studio with camera, Echo Show 16, and Alexa Photo Frame. These products enhance Alexa+ capabilities through improved…

  Wiredoriginal ↗

• 20:00Conversational AI Use Cases in Banking

  Banks deploy conversational AI to offer 24/7 support, reducing wait times and operational costs. The technology uses natural language understanding to handle routine tasks like…

  ElevenLabs BlogProducts & applicationsoriginal ↗

• 20:00ElevenLabs Launches Precise Forced Alignment Model Cue

  ElevenLabs introduces Cue, a character-level forced alignment model that aligns audio with sub-millisecond precision across all supported languages. It processes one-hour…

  ElevenLabs BlogProducts & applicationsoriginal ↗

• 18:32FluxVLA Engine V2 Technical Report Released

  FluxVLA Engine V2 introduces online reinforcement learning and expands evaluation across multiple simulators and real robots. Model success rates improve by up to 15.4 percentage…

  量子位original ↗

• 18:00Three Paths for AI in Corporate Org Charts

  OpenAI authors propose three organizational models for AI integration: beneath, beside, and between humans. In each, AI takes on tasks at different levels without full automation…

  OpenAI Newsoriginal ↗

• 13:31Australia to ban AI chatbots creating emotional intimacy

  Australia proposes banning AI chatbots that simulate emotional intimacy, especially those fostering children's dependence on non-human companions. The Digital Duty of Care Bill…

  Guardian TechPolicy & regulationoriginal ↗

• 12:00HHS Issues Rule on Health Data Technology Standards

  The U.S. Department of Health and Human Services issued a final rule on October 28, 2026, updating health IT standards and certification requirements. It takes effect on December…

  Federal RegisterPolicy & regulationoriginal ↗

• 08:00Muse Now Available in Singapore

  Meta has launched Muse in Singapore, a personal AI agent that integrates with local apps to assist daily tasks. It connects to services like Shopee, Grab, and Singapore Airlines…

  Meta NewsroomProducts & applicationsoriginal ↗

• 08:00Google launches LiteRT for faster Android inference

  Google introduces LiteRT in Play Services for faster on-device ML inference, with up to 1.4x GPU speed improvement via ML Drift and CompiledModel API. Apps reduce size by ~8 MB as…

  Google AI ChangelogProducts & applicationsoriginal ↗

• 08:00Google releases Agents CLI eval loop for agent quality hill-climbing

  Google introduces an Agents CLI evaluation loop for iterative agent quality improvement. It starts with a small test suite of four scenarios, achieving a mean score of 4.0 with…

  Google AI ChangelogProducts & applicationsoriginal ↗

• 08:00Anthropic expands AI for Science to institutions

  Anthropic has expanded its AI for Science program to 20 global scientific foundations and institutions, providing free Claude credits to researchers. The initiative supports…

  Anthropic Newsoriginal ↗

• 06:44Gemini Notebook adds built-in audio recorder to Android and iOS

  Gemini Notebook has added an in-app audio recorder to both Android and iOS. Users can start recording via a microphone icon on the home screen or Sources tab, with recordings…

  9to5GoogleProducts & applicationsoriginal ↗

• 03:30Oracle AI Database@Google Cloud Adds Gemini Integration and Agentic Capabilities

  Oracle AI Database@Google Cloud integrates with Gemini and enables agentic workflows through secure, standardized SQL access and unified data governance. Features include remote…

  Google CloudProducts & applicationsoriginal ↗

• 03:00Databricks Launches Vital Skills for Genie Code in Healthcare & Life Sciences

  Databricks introduces Vital Skills, an open-source agent skill library for healthcare and life sciences, to enhance Genie Code's domain expertise. Tests on BiomniBench show…

  Databricks BlogProducts & applicationsoriginal ↗

• 01:35Google proposes 'sleep framework' for LLM continual learning

  Google introduces a 'sleep framework' enabling LLMs to continuously learn and retain knowledge through periodic offline consolidation. The model alternates between active input…

  Google Research BlogAI for Scienceoriginal ↗

• 01:16Liquid AI Launches LFM3 Multimodal Foundation Model

  Liquid AI launches LFM3, a new generation of multimodal foundation models with unified backbone for text, vision, audio, and decision tasks. All models are 2B-parameter and…

  Hugging Face BlogProducts & applicationsoriginal ↗

• 01:08OpenAI Claims Models Refute ETH

  OpenAI claims its models have refuted the Exponential Time Hypothesis (ETH), a foundational assumption in computational complexity. The lab released solutions across three…

  TechCrunch AIFunding & companiesoriginal ↗

• 01:00NVIDIA RTX Spark Enables AI Rendering and Characters in Games

  NVIDIA RTX Spark introduces AI rendering and intelligent character technologies for gaming on slim laptops and compact desktops. DLSS 5 with 3D-Guided Neural Rendering enhances…

  NVIDIAChips & computeoriginal ↗

• 00:50Exowatt beats frontier models with SageMaker serverless customization

  Exowatt uses SageMaker serverless model customization to build edge and cloud-specific models for alarm triage and energy market analysis, achieving higher accuracy than frontier…

  AWS Machine Learning BlogProducts & applicationsoriginal ↗

• 00:30Google launches Genie 4 world model with real-time audio and scene editing

  Google DeepMind released Genie 4 as a research preview, enabling photorealistic worlds with real-time audio and dynamic scene editing. The model maintains consistency for up to…

  TechCrunch AIFunding & companiesoriginal ↗

• 00:14Firefox Smart Window adds AI form filling and page monitoring

  Mozilla updated Firefox's Smart Window with AI-powered form filling from open tabs and webpage monitoring. The feature uses data from active pages to suggest text inputs…

  The Verge AIAI × humanities & societyoriginal ↗

• 00:12LLMs' reliability in judging research novelty assessed

  A study evaluates LLMs' reliability in judging research novelty using human-authored ideas from ICLR 2026 reviews. Results show significant variation across models due to prompt…

  Hugging Face BlogProducts & applicationsoriginal ↗

• 00:00NVIDIA and Oracle Expand Vera Rubin to Multi-Gigawatt Scale on OCI

  NVIDIA and Oracle expand NVIDIA Vera Rubin to multi-gigawatt scale on OCI, enabling AI training, inference, and agent workloads. The platform supports independent CPU-GPU scaling…

  NVIDIA NewsroomChips & computeoriginal ↗

• 00:00NVIDIA and Pleias Release France and Belgium Personas for Local Model Adaptation

  NVIDIA and Pleias co-developed region-specific persona datasets for France and Belgium, grounded in demographic, geographic, linguistic, and cultural data. These are used to…

  NVIDIAChips & computeoriginal ↗