This issue collects 32 AI updates from 2026-07-17, led by NVIDIA Nemotron 3 Embed Ranks 1 Overall on RTEB, Advancing Agentic Retrieval, Google AI Mode Now Connects to Third Party Apps, Enabling Task Completion in Search, Google Rebrands NotebookLM as Gemini Notebook, Accelerating AI Product Ecosystem Integration, Google Vids Launches Gemini Omni Multimodal AI and Personal Avatars for Video Creation.
NVIDIA released its Nemotron 3 Embed model, which achieved the top overall score on the RTEB retrieval benchmark.The model is optimized for retrieval tasks in agent workflows, aiming to improve AI agents' ability to gather information in complex scenarios.
Midu has unveiled DataQ, a data analysis and insights intelligent agent, at the 2026 World Artificial Intelligence Conference (WAIC) in Shanghai.The agent enables users to query data, detect trends, and generate reports through natural language interaction, lowering the barrier to professional data analysis.
Baidu's universal intelligent agent 'Baidu Dazi' has been awarded the prestigious 'Treasure of the Hall' honor at WAIC 2026, with fewer than 10 products selected annually.Baidu is showcasing its full agent ecosystem at the conference, including Miaoda, Yijing, GenFlow 4.0, and the DAA (Daily Active Agents) white paper co released with IDC.
StepFun (阶跃星辰) has unveiled the STEPX Neo, a native AI agent smartphone that integrates large language model capabilities at the operating system level.The device has entered the production scheduling phase, marking a milestone for AI native hardware in China.
WAIC 2026商汤大装置发布算电协同Agent，单位电力成本Token产出提升80%.According to qbitai.com, WAIC 2026商汤大装置发布算电协同Agent，单位电力成本Token产出提升80%.
Anthropic's Claude has integrated with 1Password, enabling the AI assistant to securely log into websites and perform tasks on behalf of users without directly accessing their passwords.The move extends Claude's agentic capabilities to authenticated online services.
At the 2026 World Artificial Intelligence Conference in Shanghai, the personalization of AI agents became a hot topic of discussion.Beidian Shuzhi showcased real world AI applications for public welfare, demonstrating how AI is being deployed in everyday life.
Ledger, the company best known for cryptocurrency hardware wallets, has introduced an AI Agent Stack designed to connect artificial intelligence tools.The move marks a significant expansion from crypto security hardware into AI infrastructure.
Google announced that AI Mode now supports securely linking third party apps, allowing users to interact with services directly in search results.This update transforms Google Search from an information tool into a task execution platform.
Google has officially renamed its AI note taking tool NotebookLM to Gemini Notebook, strengthening integration with the Gemini ecosystem.The rebranding aims to improve product discoverability and user experience.
Google Workspace's video creation tool Google Vids received two major updates: Gemini Omni multimodal capabilities and personal AI avatar features.Users can now create, edit, and star in video content more easily than ever.
Google's next generation flagship AI model, Gemini 3.5 Pro, has been delayed due to internal bureaucracy and the model failing to meet performance targets, according to multiple reports.Alphabet shares fell on the news, reflecting market concerns about Google's competitive position in the AI race.
Roblox has introduced a new AI powered feature called “Build” in its mobile app that generates basic games from a single text prompt.The move dramatically lowers the barrier to game creation, allowing any user to transform their ideas into interactive experiences without coding skills.
Honda and Acura are rolling out Google Gemini AI assistant to nine Honda and three Acura models, replacing the old Google Assistant with a conversational AI that understands layered questions and retains context.Drivers can ask for directions, news summaries, trip planning, and more in natural free flowing conversation.
Apple's AI system has completed algorithm filing in China, clearing a key regulatory hurdle for launching AI features to Chinese users.Reports indicate Apple has formed differentiated AI partnerships with Alibaba's Qwen and Baidu.
OpenAI has unveiled a set of safety measures for teenage ChatGPT users, including age appropriate protections, learning tools, parental controls, and expert partnerships.The initiative aims to balance teens' access to AI with responsible safety guardrails.
WeRide has unveiled WITT (World Intelligence Toward Truth), a physical AI foundation model that introduces the concept of "minimum physical fact units" to transform autonomous driving video into structured, searchable data.The company claims WITT can boost data processing efficiency by up to 200x, handling 10,000 minutes of driving video per day on a single GPU at one third the factual error rate of general purpose large models.
During WAIC 2026, Shanghai's Jing'an district launched the Connect Jingan: Scene Emergence event and released an AI application industry map.The map provides a panoramic overview of how AI technologies are deployed across various real world application scenarios.
Moonshot AI's release of the Kimi K3 model has triggered a sharp decline in AI and semiconductor stocks.The market reaction, reported by Crypto Briefing, underscores the high sensitivity of investors to shifts in the AI model competitive landscape.
A survey of 157 enterprises reveals that organizations are granting AI agents more autonomy while trusting the evaluations meant to gate that autonomy less.Half have shipped an agent that passed internal evaluations only to fail in customer production environments.
The AI compute gap: Enterprises are buying infrastructure faster than they can measure what it costs.Across 107 enterprises, AI infrastructure spending is accelerating well ahead of the ability to see or steer its economics.Most organizations run their AI on a familiar base of hyperscalers and model provider APIs, yet the next dollar is aimed at specialized.
A VentureBeat survey of 107 enterprises found that more than half have already experienced a confirmed AI agent security incident or near miss.Only about one third give every agent its own scoped identity, and most agents still share credentials, exposing organizations to lateral movement risks.
A VentureBeat survey of 101 enterprises found that RAG has become the default context source for enterprise AI, but most organizations have already watched their agents produce confident wrong answers traced to missing or inconsistent context.A governed semantic layer is emerging as the next critical infrastructure investment.
IDC's latest report shows Alibaba's Qoder captured 47.6% of China's AI coding market revenue in 2025, exceeding the combined share of the second through fifth competitors.The agent powered coding platform, launched less than a year ago, has reached 5 million users by offering a comprehensive product line spanning IDE, cloud agent, CLI, and digital employee tools.
Chinese AI lab DeepSeek has completed a $7.4 billion funding round, positioning it among the most well capitalized AI startups globally.The massive raise marks the ascension of a Chinese domestic AI lab into the global first tier of AI companies.
Nokia and NVIDIA have launched the industry's first GPU based AI RAN solution, aiming to double wireless spectrum capacity.The partnership embeds NVIDIA's AI computing directly into Nokia's radio access network infrastructure, marking a shift toward AI native telecom architecture.
Why the first GPU financiers are turning to inference chips in a $400 million deal.A $400 million chip backed loan points to the next wave of AI infrastructure deals.
Damao Technology launched its Computing Electricity Coordination 2.0 platform at WAIC 2026, featuring a dual engine architecture combining inference intelligence and game theoretic optimization.The system aims to reduce per kWh electricity costs by 10 20% for GW scale AI data centers.
Sakana AI has partnered with NVIDIA to integrate Nemotron 3 Ultra into its Fugu multi model orchestration system.The move reinforces the thesis that intelligent routing of open weight models can outperform reliance on any single proprietary model across diverse tasks.
Moonshot AI has officially released Kimi K3, which it claims is the world's largest open source model, featuring vision understanding capabilities and a 1 million token context window.The release represents a significant milestone in open source large language model development.
Satya Nadella publicly criticized Anthropic's data usage policy, calling it nonsensical.The remark highlights growing tensions between the two AI companies over how customer data should be used for model training and improvement.
The Australian government has unveiled what it describes as a world first national AI regulatory framework, taking an active stance on AI governance under the theme “On our terms.” The framework addresses multiple aspects of AI development, deployment, and use, according to analysis from law firm MinterEllisonRuddWatts.
