AI News

Italian regulator suspends Meta's policy restricting third-party AI access to WhatsApp Business API, citing abuse of market dominance and harm to AI competition and consumers. This follows EU investigations, signaling systematic antitrust scrutiny of Meta's AI strategy in Europe.....

Loomi launched the world's first AI content creation IDE, aiming to address pain points faced by creators such as low efficiency and difficulty in content adaptation. By leveraging AI technology, it optimizes the creative process and enhances content output efficiency.

In 2025's competitive self-media landscape, creators face core anxiety over lack of content feedback. While viral hits are often attributed to luck, the algorithm era demands a shift from intuitive creation to industrialized production, with AI tools driving a paradigm from generation to research.....

Lenovo to launch 'AI Super Agent' at CES, a global strategic application rivaling ByteDance's 'Doubao Phone Assistant'. It features advanced functionality and ecosystem synergy, surpassing basic voice assistants, marking Lenovo's full-scale push into AI applications.....

HodlHer secures $1.5M in strategic funding from investors including Chain Capital. The funds will advance its core OS, HodlOS, focusing on emotional interaction, multi-agent systems, and on-chain agent economics to deploy the next-gen Web3 OS, enabling smarter agents to understand user emotions and navigate complex Web3 environments.....

OpenAI partners with Broadcom to develop custom AI inference chips, set for 2026 deployment, aiming to deploy 10 GW of computing power over five years to move beyond GPU reliance and reshape the AI economy.....

Top AI models like ChatGPT and Claude overestimate human rationality in decision-making, assuming highly logical behavior. Tests using the 'Keynesian Beauty Contest' reveal their failure to predict actual human choices, highlighting a gap between AI assumptions and real-world human decisions.....

Beijing's open-source ecosystem plan aims to cultivate 10 globally influential projects (including 5 leading ones) and 30 domestic star projects (over 1/3 AI-related) by 2028, while promoting large-scale application of 100 industry models to build a world-leading open-source innovation hub.....

NVIDIA releases a guide for fine-tuning large models, lowering technical barriers to enable developers to efficiently customize models on consumer-grade devices. It details using the Unsloth framework across NVIDIA hardware for optimized, professional-grade fine-tuning.....

OpenAI introduces 'formatting blocks' for ChatGPT, automatically adjusting UI for tasks like emails and blogs, enhancing content creation with a mini-editor toolbar for rich text editing.....

OpenAI is developing 'Skills', a modular and reusable AI feature set to launch in 2026, transforming user-AI collaboration by shifting from custom assistants to cross-platform 'capability modules'.....

Alibaba Cloud's Tongyi Lab open-sources Qwen-Image-Edit-2511, enhancing editing consistency and visual stability to reduce 'slight drift' issues, offering developers a more precise and reliable controllable generation tool.....

Nvidia secures non-exclusive AI chip tech license from Groq for ~$20B, absorbing its core team to sidestep regulatory hurdles, a move seen as strategic defense against Google's TPU competition.....

BYD partners with Volcano Engine to integrate Doubao AI model into DiLink smart cockpit across all five brands, achieving the world's largest AI-enabled smart cabin upgrade for enhanced interactive experiences beyond traditional voice assistants.....

Cursor's CEO warns that over-reliance on AI for 'ambient programming' risks software engineering issues by preventing developers from examining underlying code.....

China achieves breakthroughs in integrated circuits and AI, with over 700 devices using HarmonyOS, generative AI models approved, and embodied intelligence advancing, accelerating the shift to a tech powerhouse and core technology self-reliance.....

The 2026 Beijing Yizhuang Humanoid Robot Half Marathon on April 19 features a 'human-robot co-run' format, with separate lanes for competitors. It evaluates robot autonomy, human-like gait, and endurance through autonomous navigation and remote-control categories, incentivizing full autonomy. A 'Baturu' challenge simulating emergency rescue will also be held.....

DingTalk and Xiaoice jointly launched AI receptionist Hi1, using 'AI hardware + digital employee' to redefine corporate reception with interactive capabilities, addressing issues like repetitive tasks and inconsistent service standards.....

OpenAI seeks $100 billion funding to fuel growth, potentially valuing it at $830 billion. Early-stage talks aim for Q1 completion, though waning market enthusiasm could dampen investor interest.....

InstanceAssemble is an open-source controllable image generation framework by Xiaohongshu, designed for high-density, multi-object, and complex spatial tasks. It uses cascade modeling and Assemble-Attention to enhance spatial alignment and semantic consistency with minimal added parameters, offering industrial-grade solutions for e-commerce and design.....