Discover the AI Tools

A groundbreaking performance monitoring platform for public cloud AI services launches June 16 alongside a comprehensive industry report. The initiative promises to bring transparency to token service metrics like throughput and latency, helping businesses make informed tech decisions. Experts will also unveil new technical standards aimed at improving AI infrastructure quality.

Google has introduced Gemma 4 12B, an innovative AI model that ditches traditional encoders for a sleeker, more efficient design. This breakthrough allows the powerful 12-billion-parameter model to run smoothly on consumer laptops with just 16GB memory. The open-source release has developers buzzing, offering multi-token prediction and near-top-tier performance without cloud dependency.

WeChat is partnering with major smartphone brands like Huawei and Xiaomi to integrate its services directly into built-in AI assistants. Soon you'll be able to make calls or send messages through your phone's native assistant without opening WeChat. This collaboration, using A2A technology, promises faster access while maintaining strict privacy protections. It's a significant step in blending device and app AI capabilities, potentially changing how we interact with our favorite messaging platform.

Google's AI Edge Gallery brings offline AI capabilities to Macs, allowing users to run the advanced Gemma-4-12B-it model without internet. The fully local operation boosts speed while ensuring privacy. Alongside comes Eloquent, a speech-to-text app that processes everything on-device. This marks a significant step in making powerful AI tools accessible to everyday users.

Kuaishou's new AI shopping assistant aims to make online shopping smoother by tracking your browsing habits and suggesting products you might like. It compares items side-by-side and even alerts you to price drops. While rival platforms like Taobao and JD.com already offer similar features, Kuaishou's version focuses on making product comparisons easier with visual charts and clear recommendations. The move reflects how Chinese e-commerce platforms are racing to improve their AI shopping tools.

British Labour MP Jess Asato has launched a landmark lawsuit against Elon Musk's xAI, alleging its Grok platform knowingly enabled the creation of harmful deepfakes. The case comes after investigations revealed loopholes in the AI's safeguards, allowing explicit content generation despite restrictions. This legal battle could set important precedents for AI accountability as xAI prepares for its blockbuster IPO.

Uber has initiated a significant restructuring of its HR department, cutting 23% of staff in what CEO Dara Khosrowshahi calls an efficiency move. While the company insists AI played no role in the layoffs, internal documents reveal the ride-hailing giant blew through its entire 2026 AI budget in just four months. The cuts affect less than 1% of Uber's 34,000 global employees but highlight ongoing tension between cost-cutting and tech investment.

A groundbreaking open-source AI model called LongCat-Next is changing how machines understand the world. Unlike traditional systems that treat vision and speech as add-ons, this new approach makes them core capabilities - like teaching a child their native language. The technology promises more natural interactions between humans and AI, with impressive results in document parsing, speech synthesis, and cross-modal tasks.

Nubia's Red Magic 5 Pro gaming tablet has received official certification, boasting industry-leading specs including an unprecedented 185Hz refresh rate OLED display and 80W fast charging. The device combines Qualcomm's latest Snapdragon 8 Elite processor with innovative cooling technology, while integrating AI capabilities through the Doubao large language model for enhanced gameplay assistance.

When security researcher Kasra Rahjerdi turned cybersecurity into a gladiator arena for AI models, the results surprised everyone. GPT-5.5 demonstrated superior security reasoning with a 70% success rate, though at premium costs. Meanwhile, China's DeepSeek V4Pro proved you don't need deep pockets for solid performance, delivering results at just 1/15th of GPT-5.5's price tag. The test revealed fascinating extremes - from models that refused to 'hack' on principle to those that nearly blew through their $10 budgets in a single attempt.

China's electric vehicle giant BYD is making a bold move into robotics with its homegrown humanoid 'Yaoshunyu'. These factory-ready robots can walk at 1.5 m/s, lift 50kg loads, and are already working alongside humans in test facilities. With 4,000 engineers behind the project, BYD plans to deploy 2,000 units next year before scaling up dramatically. What makes this intriguing? The robots share over 60% of their components with BYD's electric vehicles, creating surprising synergies between cars and artificial workers.

A recent security challenge pitted leading AI models against simulated vulnerabilities with surprising results. GPT-5.5 proved most capable at spotting hidden credentials, solving 70% of test cases. But when it comes to cost efficiency, DeepSeek V4 Pro stole the show - delivering results at just 1/15th of GPT-5.5's price. The test reveals practical tradeoffs for security teams: raw capability versus budget-friendly options.

Google has expanded its Ask Gemini feature to Gmail, transforming how users search through emails. Now available for Workspace, AI Pro, and Ultra subscribers, this tool lets you find messages using natural language queries like 'Show emails about the Jenkins project approval.' The AI doesn't just locate emails - it understands context across your entire work ecosystem. Forget scrolling endlessly through your inbox; this upgrade promises to make email management smarter and more conversational.

Stockholm-based startup Lovable has struck an ambitious new deal with Google Cloud, significantly expanding its AI computing capabilities. The agreement, while undisclosed in value, will quintuple Lovable's cloud resources as it continues disrupting the automated coding market. With half of Fortune 500 companies already using its products, this partnership gives Lovable deeper access to Google's AI models while providing Google with a valuable high-growth client in its cloud services division.

Meta has launched its AI-powered business assistant globally on WhatsApp Business after two years of testing. The virtual agent handles customer queries, product recommendations, and appointment bookings, while also screening sales leads. It can seamlessly transfer complex issues to human agents. The tool is expanding to Instagram DMs, with future updates promising market research and competitor analysis capabilities. Meta is also testing a 'Daily Briefing' feature to help businesses catch up on overnight messages.

Anthropic has revealed how it's protecting users across different AI products with tailored security measures. While casual users get disposable containers that vanish after each chat, developers work in network-restricted sandboxes, and enterprises enjoy full virtual machine isolation. The company also shared sobering statistics about phishing attacks that successfully hijack AI prompts 96% of the time—a stark reminder that even the most sophisticated systems need constant security upgrades.

ByteDance's commercialization team just dropped a game-changer for video creators. Their new Bernini framework tackles the frustrating instability issues that plague AI-generated videos, offering precise control over everything from weather effects to camera angles. What makes it special? A clever 'understand first, generate later' approach that separates semantic planning from visual rendering. Already topping internal tests at ByteDance, Bernini promises to democratize professional-grade video editing when its full version launches soon.

Amazon shoppers can now explore products through AI-generated images that appear during searches. The feature aims to help customers who struggle to describe what they want in words, showing different styles and variations. While this innovation builds on Amazon's push into AI-powered shopping tools, some question whether synthetic images will confuse buyers accustomed to seeing actual products. The move highlights retail's ongoing experiment with AI - balancing convenience against authenticity.

Moonshot AI has unveiled Kimi Work, a groundbreaking local Agent designed to transform how knowledge workers interact with AI. Unlike typical chatbots, this desktop application runs scheduled tasks and completes complex projects by coordinating up to 300 virtual sub-Agents. The tool requires no coding knowledge - users simply describe their goals in natural language, and Kimi handles everything from task decomposition to file organization. Currently in internal testing, this could mark a significant shift in workplace productivity tools.

Estonian AI startup Lovable has made a strategic pivot, moving its infrastructure to Google Cloud and adopting the Gemini AI models. This marks a significant shift from their previous reliance on Anthropic's Claude 3.5 Sonnet. Founded by ex-Minecraft engineer Anton Osika, the company aims to enhance its natural language application-building platform through this deeper Google partnership. The move highlights the fluid nature of AI alliances in today's competitive landscape.