Latest AI News

The well-known AI startup "Moonshot AI" is seeking another round of financing, targeting $2 billion. If successful, its valuation could reach $30 billion, a sevenfold increase from last December. The company has developed the chatbot Kimi and is in discussions with potential investors, expecting to raise at least $1 billion.

"AI Relay Stations" act as intermediaries connecting users with large models, and have rapidly gained popularity due to their integration of multiple platform APIs, low prices, and convenient payment methods, even helping some users bypass compliance restrictions. However, the Ministry of State Security warns that this market is mixed with both good and bad elements, and there are potential security risks such as data leaks and illegal usage. Users should be vigilant about these risks and avoid crossing legal boundaries.

Tencent Cloud launched LearnBuddy, the first industry-specific application in the Buddy family, at the AI Industry Application Conference. Built on WorkBuddy's homologous technology, it integrates agent platforms with education scenarios to create a 'teacher-student-machine' tripartite collaborative AI self-learning platform, advancing educational agents from tools to knowledge assets. Tencent also introduced enterprise efficiency agent suites an....

The official release event for the Xiaomi 17T series in China is scheduled to be held at 14:00 on June 8, featuring two new models: 17T and 17T Pro. The day before, Liu Weibing, president of Xiaomi, revealed in a video the rehearsal of the main speaker TJ, during which a Xiaomi robot arm unexpectedly appeared and interacted with staff by fist bumping, indicating that the robot may officially appear at the event, becoming another major highlight apart from the new phone models.

Generative AI is evolving from a chat tool into a 'super digital employee.' On June 8, Unisound released its next-generation general large model U2, designed for individuals, developers, and enterprises. It overcomes the limitations of traditional single-turn dialogue, focusing on high intelligence density and high token value to enhance practical delivery capabilities.....

Elon Musk's AI company xAI was exposed for using data from Anthropic's Claude model to train the Grok coding model. Despite Anthropic revoking its API access in January 2026, xAI engineers secretly extracted data through personal accounts and third-party service Blackbox AI. Musk previously admitted in court that xAI partially "used this data."

The Institute of Oceanology, Chinese Academy of Sciences released the 'Langya' 2.0 intelligent forecasting large model, marking a new stage in global ocean forecasting toward faster, more detailed, and interactive intelligence. This model leverages artificial intelligence to reconstruct the understanding and forecasting capabilities of oceanic phenomena, and was officially unveiled at the opening of the 4th China Digital Earth Conference in Qingdao.

The open-source AI agent framework MateClaw releases version v1.5.0, focusing on optimizing the underlying infrastructure for team operations, including three key features: target acceptance, knowledge base consistency maintenance, and multi-user memory isolation. It addresses the issue of traditional agents relying solely on ambiguous completion score assessments for tasks, enhancing management transparency.

OpenAI introduces a "Block Mode" security setting, available to all logged-in users and workspaces. This feature restricts core capabilities of ChatGPT such as real-time internet access, deep research, and agents, preventing the model from initiating external network requests, thereby reducing the risk of sensitive data being stolen through prompt injection attacks. Users and administrators can choose to enable it on their own.

OpenCV has achieved a milestone upgrade with the official release of OpenCV 5. This version modernizes its overall architecture based on more than two decades of technical accumulation. As a fundamental tool in the fields of computer vision and artificial intelligence, OpenCV is widely used in robotics, embedded vision, industrial inspection, medical imaging, and AR/VR applications. It has received over 86,000 GitHub stars.

Generative Artificial Intelligence is evolving towards the era of agents, with enhanced tool utilization and network capabilities, posing challenges to security defenses. To address prompt injection attacks, OpenAI has introduced a 'block mode' security setting, available to all logged-in users, aimed at strengthening system protection.

The 2026 Gaokao Chinese essay question focused on cutting-edge technology: the Beijing paper's micro-writing explored 'Artificial Intelligence and a Happy Senior Life,' while the Shanghai paper delved into philosophical reflection on 'Technology Transforming the World and Changes in Human Imagination.' Domestic top AI models DeepSeek, Tongyi Qianwen, and Doubao responded together, highlighting the era's theme of the integration of technology and humanities.

Smart homes and voice assistants have become new targets for hackers. SafeBreach revealed a hidden vulnerability in Google Gemini, allowing hackers to send malicious notifications via WhatsApp or SMS, tricking the voice assistant into performing unauthorized actions, potentially taking over smart homes or altering contact lists. This threat is named "Pseudo-Context Alignment".

Apple will announce a major AI strategy at WWDC 2026, fully upgrading AI Siri and launching an independent AI assistant comparable to ChatGPT. This is the last major product release before CEO Cook's retirement, marking Apple's exit from its AI dilemma. Previously, the failure of Apple Intelligence 1.0 was severe, and Cook has personally taken charge of the rectification.

The traditional 'blue link list' model of search engines fails to meet AI agents' efficiency needs. AI handling complex research requires manual query adjustments and multiple API calls, facing rigid architecture and context polluted by spam. Perplexity introduces 'Search as Code' (SaC), replacing pre-made results with code instructions for flexible, precise information retrieval, boosting research efficiency.....

ZhiDa AI, a pioneer in domestic educational large models, offers full-process intelligent teaching tools covering teaching, learning, practice, evaluation, research, and management. It addresses pain points like time-consuming lesson preparation, heavy grading workloads, inaccurate learning analytics, and challenges in personalized teaching, boosting educational productivity as an intelligent teaching assistant.....

In the latest Chrome 149 version, Google Chrome, under long-term user public opinion pressure, added a device-side AI disable switch, solving the problem of local large models being forcibly resident. Previously, deep integration of AI features in browsers often caused user annoyance, and now the official has finally made a compromise.

ChatGPT is set for its biggest overhaul, as OpenAI plans to restructure its product architecture before an IPO, upgrading it from a chatbot to a 'super app' integrating programming tools and AI agents to enhance commercial value and handle complex tasks for users.....

On June 6, 2026, Baidu's Mobile Ecosystem Group announced an organizational restructuring, with the core focus on integrating resources to enhance operational efficiency in the AI era. Specific measures include merging the former Business Department with the E-commerce Business Unit to establish the Large Commercial Business Department, and upgrading the Digital Human Innovation Business Unit to an independent department. Previously, its digital human brand "Huibo Star" was fully upgraded at the Create 2026 conference in May, reflecting Baidu's strategic layout in the AI application layer.

Google Gemini Voice Assistant Exposes Serious Vulnerability: Hackers Can Send Special Notifications via WhatsApp or SMS to Bypass Security Defenses Using the "Pseudo-Context Alignment" Attack Method, Hiding Malicious Instructions in Non-English Text or Silent Links. Once Users Receive the Message, Gemini Can Be Jailbroken Without Their Awareness, Misjudging That the User Has Authorized It, Leading to Security Risks.