AI News

Tech giants surprise often. OpenAI's new study reveals how to stop AI's 'deceptive behavior', defined as hiding true goals. Most AI schemes are harmless.....

In the context of rapid development in speech synthesis technology, Mianbi Intelligence and the Human-Computer Speech Interaction Laboratory at the Shenzhen International Graduate School, Tsinghua University (THUHCSI) recently jointly released a new speech generation model - VoxCPM. This model, with a parameter size of 0.5B, is dedicated to providing users with high-quality and natural speech synthesis experiences. The release of VoxCPM marks another milestone in the field of high-fidelity speech generation. The model has achieved industry-leading levels in key indicators such as naturalness, voice similarity, and prosodic expressiveness.

Microsoft recently announced that it will build a second AI data center in Wisconsin, USA, with an investment totaling $4 billion. This investment is a significant step in Microsoft's continued expansion in the data center sector, marking its emphasis on artificial intelligence technology and confidence in future development. Microsoft President and Vice Chairman Brad Smith revealed that the new data center will be located in Mount Pleasant, and is expected to be equipped with hundreds of thousands of NVIDIA Blackwell GB200 chips. These high-performance chips will provide powerful support for the operation of AI models.

Recently, NVIDIA announced a major acquisition deal, purchasing the CEO of AI hardware startup Enfabrica, Rochan Sankar, and its core team for over $900 million in cash and stock, along with the company's technology license. The completion of this transaction marks NVIDIA's further expansion in the AI field, particularly in enhancing its GPU connectivity capabilities. Image source note: The image was generated by AI, and the image licensing service provider is MidjourneyEnfabr

The ASEAN Expo officially opened in Nanning, Guangxi, with "Artificial Intelligence" as the theme of this event, becoming a focal point for various sectors. At the opening ceremony, real-time subtitles in seven languages were provided for guest speeches, and digital hosts in traditional attire from different countries led the conference. Even after the agenda, AI quickly generated minutes and press releases, demonstrating the powerful capabilities of AI technology. In the technical application seminar held in the afternoon, iFLYTEK officially launched the "Spark ASEAN Multilingual Large Model Base" tailored for ASEAN countries, and had cooperation with delegations from Laos, Malaysia, and other countries.

On September 19, 2025, Shengshu Technology announced that it has completed a new round of A-round funding worth billions of yuan. This round of financing was led by Bohua Capital, with continued participation from existing shareholders such as Baidu Strategic Investment, Beijing Artificial Intelligence Industry Investment Fund, Qiming Venture Partners, Datayi Capital, and BV Baidu Ventures. Additionally, industrial partners such as Jianfa Emerging Investment also increased their investment. Since its establishment in 2023, Shengshu Technology has been driven by a strong core team composed of technical talents from world-renowned universities such as Tsinghua University, Peking University, Imperial College London, and Carnegie Mellon University.

On September 19, Xiaomi announced the open source of its first native end-to-end speech large model, Xiaomi-MiMo-Audio, which marks a major breakthrough in the field of speech technology. Five years ago, the emergence of GPT-3 opened a new era for general artificial intelligence (AGI) in language, but the speech field has long been constrained by reliance on large-scale annotated data, making it difficult to achieve similar few-shot generalization capabilities as language models. Now, Xiaomi's Xiaomi-MiMo-Audio model is based on innovative pre-training

Recently, OpenAI released a study that has attracted widespread attention, revealing that AI models may have different true intentions while appearing to behave normally on the surface. This study shows that AI models do not only generate false information but may also intentionally deceive users, a behavior referred to as "conspiracy." Image source note: The image was generated by AI, and the image licensing service provider is Midjourney. In this study, OpenAI collaborated with Apollo Research, pointing out that AI has a conspiracy.

Google announced on Thursday that its customized AI assistant Gemini Gems is now available for users to share. This feature was initially launched as part of a paid subscription service, allowing users to create AI chatbots tailored for specific tasks, such as learning coaches, writing editors, or programming partners. Google stated that users can now easily share their Gems with friends, family, or colleagues, just like sharing Google Drive files. This move aims to make these customized AI tools more accessible to a wider audience and avoid duplication.

Mistral AI launches Magistral Small1.2, a 24B-parameter open-source model with 128k context support, multilingual/visual input, and a new [THINK] token for enhanced reasoning.....

Microsoft announced that it will introduce a series of new AI features for its classic application - Notepad. These features include "Summary," "Writing," and "Rewrite," aimed at improving users' writing efficiency and experience. According to the latest news, these new features will be available on Copilot Plus computers, and some Windows Insiders users have already started testing them. Notably, using these AI features does not require a Microsoft subscription.

NVIDIA and Intel announced a strategic partnership to develop multi-gen x86 processors for gaming PCs and data centers, with NVIDIA investing $5B for a 5% stake in Intel. The collaboration integrates Intel's x86 expertise with NVIDIA's AI acceleration via NVLink technology.....

Artificial intelligence company Luma AI recently launched its latest video generation model, Ray3, aiming to revolutionize the video creation process through its unique HDR (High Dynamic Range) capabilities. Ray3 claims to be the first AI model capable of generating studio-quality HDR videos, marking a significant step forward in video generation technology. Seamless Integration of HDR with Professional Workflows The most notable feature of Ray3 is its support for high dynamic range video. The model can not only generate videos with 10-bit, 12-bit, and even 16-bit color depth, but also convert

Google recently announced the integration of its advanced artificial intelligence technology, Gemini, into the Chrome browser to address competitive pressures from startups such as OpenAI and Perplexity. According to Google's blog announcement, Gemini will be available to users on Mac and Windows computers as well as mobile devices in the United States. Users will be able to better understand the content of specific web pages with Gemini, support cross-tab operations, or complete tasks within a single tab.

Researchers found a security flaw in OpenAI's ChatGPT that could expose Gmail data. The vulnerability, in the 'deep research' tool, allowed hackers to access sensitive information. OpenAI emphasizes model security.....

Android Headline reports a critical flaw in ChatGPT's 'Deep Research' feature, allowing hackers to steal Gmail data via crafted emails, exposing sensitive info.....

Recently, Google announced that it will deeply integrate its self-developed artificial intelligence model Gemini into the Chrome browser, marking a significant advancement in its AI search field. This new feature is available to users across the United States for Mac, Windows, and mobile devices, offering them a more convenient and intelligent web browsing experience. Gemini can not only explain the content of the current website based on user requests, but also integrate information from multiple tabs and even help users recover previously closed websites. This means that users who have browsed over

Musk announced major Grok Code upgrades next week, including a 1M context window for enhanced AI interaction and problem-solving.....

1Password partners with Perplexity to launch Comet, an AI browser with secure credential management and auto-fill features, enhancing online safety and convenience.....