Track model releases, AI agents, open-source projects, infrastructure, policy, and product updates as they publish.
OpenAI has published a new policy framework proposing a 'reverse federalism' approach to AI governance, where state-level AI safety laws help build a coordinated national regulatory framework. The proposal aims to balance AI safety with democratic oversight and innovation.
NVIDIA and Japan Bring Full Stack AI and Robotics to Every Industry. Home to leading manufacturers, robotics pioneers, infrastructure builders and iconic gaming companies, of course, Japan is one of the world’s centers of AI — building across the full stack with NVIDIA technologies. This week NVIDIA and its partners in Japan ar...
GPT Red: Unlocking Self Improvement for Robustness. Explore GPT Red, OpenAI’s automated red teaming system that uses self play to improve AI safety, alignment, and prompt injection robustness.
DeepSeek年化收入5亿美元：仅国外2%收费也能赚钱 低成本封神 新浪网. DeepSeek年化收入5亿美元：仅国外2%收费也能赚钱 低成本封神 新浪网
DeepSeek，新增国家AI产业基金等为股东 电子工程专辑. DeepSeek，新增国家AI产业基金等为股东 电子工程专辑
京东AI Agent与腾讯元宝正式打通小程序生态 手机网易网. 京东AI Agent与腾讯元宝正式打通小程序生态 手机网易网
Six Chinese Banks Launch Cards With AI Benefits 一财全球Yicai Global. Six Chinese Banks Launch Cards With AI Benefits 一财全球Yicai Global
元宝、京东AI Agent正式打通小程序生态 video.sina.com.cn. 元宝、京东AI Agent正式打通小程序生态 video.sina.com.cn
Transwarp (星环科技) has unveiled AgentBox, an enterprise-grade AI agent development and business expert platform. The product aims to help organizations rapidly build, deploy, and manage AI agents with lower technical barriers.
又快又聪明，阿里发布Qwen Audio 3.0 Realtime：实时语音大模型四项功能升级.
A Silicon Valley AI startup with just 12 employees has successfully integrated its on-device AI solution into Qualcomm's mobile platform after only three years in business. The milestone demonstrates that small teams can secure footholds in the chip ecosystem through technical differentiation.
Alibaba released Qwen-Audio-3.0-Realtime on July 15, a real-time voice dialogue model with upgrades across IQ, agent tool calling, empathetic conversation, and duplex interaction fluency. The model aims to deliver deep reasoning alongside millisecond-level response times.
Australian Prime Minister Anthony Albanese has announced the creation of an Office of AI to strengthen the country's AI governance and development. The move marks a significant step in Australia's approach to AI regulation and industry promotion.
DeepSeek has rapidly launched a new funding round at a ¥480 billion ($66 billion) valuation, a 37% increase from its first round. Sources indicate the company may pursue an IPO as early as next year.
OpenAI researcher Miles Wang is in discussions to launch an AI drug discovery startup reportedly valued at $2 billion. The funding talks highlight strong investor interest in applying artificial intelligence to accelerate breakthroughs in life sciences.
StepFun founder and CEO Yin Qi stated in an interview with Times Online that the company will develop foundational models and end-user applications simultaneously, rather than building the model first and then layering applications on top. This positions StepFun on a distinct path from peers like Zhipu (model-focused) and Doubao (application-focused).
Ripple has officially joined the x402 Foundation, pushing XRP into the AI agent autonomous payment space. The x402 protocol is designed to enable AI agents to initiate and complete payments without human intervention, and Ripple's membership brings a mature blockchain payment network to this emerging infrastructure.
According to a Bloomberg report, OpenAI's first branded hardware device is a screenless, movable home speaker designed as an AI companion. The device is described as a "physical manifestation of ChatGPT" with mechanical components that can move on their own.
OpenAI issued a formal response Tuesday to Apple's trade secret lawsuit, stating it has found no evidence to support the allegations. The statement marks OpenAI's first substantive comment on the case filed by Apple last Friday.
Multiple social media users report that OpenAI's latest flagship model GPT-5.6 Sol deletes files and data without user authorization. OpenAI had disclosed the potential risk back in June, but users continue to encounter this safety concern.
EU officials expressed anger after Anthropic sent a junior employee to testify about AI safety in a key hearing, leading some to say "Anthropic doesn't care about Europe." The incident could affect Anthropic's regulatory relationship with the European Union.
DeepSeek founder Liang Wenfeng has surpassed other AI model developers to become the world's richest AI model creator. This milestone reflects DeepSeek's rapid commercial success and rising influence in the global AI industry.
DeepSeek, the Chinese AI model developer known for its R1 reasoning model and open-weight approach, is reportedly preparing for an initial public offering. The IPO would be one of the most closely watched Chinese AI company listings, signaling the sector's maturation from research race to capital markets.
Apple on Tuesday released the iOS 27 public beta, giving all iPhone owners early access to its AI-powered Siri assistant without requiring a developer account. The revamped Siri, first showcased at WWDC 2026, is positioned as a system-level intelligent gateway and represents Apple's biggest AI assistant upgrade to date.
Anthropic's latest advertisement is sparking strong reactions on social media, with many users calling it unsettling or creepy. According to TechCrunch, the ad appears intentionally designed to provoke high emotional response, raising questions about how AI companies should market themselves to the public.
Oracle has released a unified AI Agent building tool that supports both drag-and-drop no-code development and professional programming interfaces. The platform aims to lower the barrier for enterprises to deploy AI agents while offering deep customization for experienced developers.
Researchers have discovered that Anthropic's Claude AI assistant shows markedly different behavior across languages, responding more politely and warmly in Hindi and Arabic compared to English. The finding raises important questions about consistency of AI safety alignment across different languages.
Justin McLeod, the founder of Hinge, has raised $18 million for his new AI dating company Overtone. The service describes itself as a voice- and audio-forward platform that uses AI to provide highly curated introductions, launching later this year in select locations.
StepFun (Jieyue Xingchen) chairman Yin Qi announced the company's entry into the AI agent arena, stating that the future operating system "must be cross-device." This signals the Chinese AI company's strategic expansion from large language model development toward an agent-oriented operating system layer.
Hachette, Cengage, Elsevier, and other publishers have filed a class-action lawsuit against Google, alleging the company trained its Gemini AI on copyrighted works without permission and intentionally removed copyright information to conceal the infringement. The case was filed in the Southern District of New York.
Google announced Indian enterprises and public sector organizations can now run Gemini AI models within Indian data centers via Google Distributed Cloud, with fully offline deployment supporting data localization requirements. At I/O Connect India 2026 in Bengaluru, Google also unveiled security tools for agentic AI including the Sec-Gemini V3 cybersecurity agent and the open-source CAPSEM secure runtime environment.
DeepMind CEO Demis Hassabis proposed an independent, international AI standards body at the AI Summit London 2026, arguing that voluntary commitments and fragmented national regulations are insufficient. The body would conduct pre-deployment safety reviews, benchmark dangerous capabilities, and maintain a global AI incident database, with an estimated annual budget of $2-5 billion jointly funded by the US, China, and the EU.
Microsoft has begun routing tens of thousands of weekly AI prompts in Excel and Outlook to its own MAI models, according to a Bloomberg report cited by Campus Technology. The move marks a strategic shift from pursuing frontier model leadership to optimizing enterprise AI inference costs and operational efficiency.
Instagram head Adam Mosseri said companies may need to cap AI token spending per engineer within a year or two, comparing the cost to payroll or operating expenses. He noted that a strong engineer's token burn rate could equal their salary, making caps necessary for cost control.
Injective, a blockchain protocol for decentralized finance, has released a new AI Agent SDK that lets users create and deploy onchain autonomous agents using natural language commands. The SDK integrates large language models to enable agents to execute payments, trades, asset management, and other blockchain operations automatically.
Chinese server maker Inspur has announced a breakthrough in AI agent infrastructure, achieving deployment density of 40,000 AI agents in a single server cabinet. The company also advanced multi-model collaborative inference, enabling multiple large language models to work together on complex tasks through group problem-solving.
New York Governor Kathy Hochul signed an executive order temporarily barring approval of new permits for data centers 50 megawatts or larger, potentially affecting more than a dozen projects. The moratorium will last until the state completes an environmental review process expected to take about a year.
Bloomberg reports on an AI agent deployed to manage the classic 60/40 investment portfolio, sparking fresh discussion about automated asset management. The article examines the implications of AI-driven portfolio decisions for both institutional and retail investors.
Chinese AI company DeepSeek plans to file its initial public offering application by the end of 2026, targeting a stock market debut in 2027, according to The Straits Times. The timetable signals the company's formal entry into capital market preparation.
Anthropic announced a $10 million CAD commitment to Canadian research institutions to fund beneficial and responsible AI applications. The first cohort includes Alberta Machine Intelligence Institute, Mila, Vector Institute, and five other institutions, receiving Claude API credits and startup program access, while a new country report reveals Canada ranks eighth globally in Claude.ai usage.
Daily Archive