AI News
The latest happenings in the AI world — launches, research, policy, and business.
45 published news items, newest first (latest dated Mar 7, 2026): launches (21), industry (7), research (6), policy (5), business (2), funding (2), partnerships (2). Each item links to its original source where one is available.
OpenAI Launches GPT-5.2 with Enhanced Reasoning
OpenAI has released GPT-5.2, featuring significant improvements in complex reasoning, mathematical proofs, and scientific analysis. The model shows a 40% improvement on graduate-level STEM benchmarks compared to GPT-5. Available through ChatGPT and the API with the same pricing tiers.
Google Gemini 3 Pro Preview Released to Developers
Google has released Gemini 3 Pro in preview, featuring a 4 million token context window and improved multimodal reasoning. The model demonstrates state-of-the-art performance on long-context benchmarks and is available through Google AI Studio and the Gemini API.
Anthropic Raises $5B Series E at $80B Valuation
Anthropic has closed a $5 billion Series E round led by Google and Salesforce Ventures, valuing the company at $80 billion. The funding will accelerate development of Claude 5 and expand enterprise AI safety research. Anthropic now has over 1,000 employees.
EU AI Act Enforcement Begins with First Major Fines
The European Union has issued its first enforcement actions under the AI Act, fining three companies a combined €45 million for deploying high-risk AI systems without required transparency measures and human oversight protocols. This marks a significant milestone in global AI regulation.
Meta Releases LLaMA 4 as Open Source
Meta has released LLaMA 4, its most capable open-source model family with variants from 8B to 400B parameters. The 70B version matches GPT-5 on most benchmarks while being freely available for commercial use. The release includes multimodal capabilities for image and video understanding.
Runway Launches Gen-4 Turbo with Real-Time Video Generation
Runway has released Gen-4 Turbo, capable of generating 10-second video clips in under 30 seconds. The model features improved motion consistency, better text rendering, and new camera control features. Available to all Runway subscribers at no additional cost.
Microsoft Copilot Expands to All Office 365 Personal Plans
Microsoft is rolling out Copilot AI features to all Microsoft 365 Personal subscribers at no extra cost, bringing AI-powered writing, data analysis, and presentation creation to 400+ million users. Previously, these features required a separate Copilot Pro subscription.
DeepSeek-V4 Benchmarks Show Dramatic Efficiency Gains
Chinese AI lab DeepSeek has released V4, achieving GPT-5 level performance with 10x lower inference costs through a novel sparse architecture. The model activates only 30B of its 1 trillion total parameters per query, setting new efficiency records.
Adobe Introduces Video Generation in Creative Cloud
Adobe has integrated AI video generation powered by Firefly Video into Premiere Pro and After Effects. Users can generate B-roll footage, extend clips, and create transitions using text prompts directly within their editing timeline. Available to all Creative Cloud subscribers.
Stability AI Releases Stable Diffusion 4 with Native Video
Stability AI has launched Stable Diffusion 4, the first unified model that handles both image and video generation. The open-source model supports up to 30-second video clips, improved text rendering, and ControlNet v3 for precise control. Requires 16GB VRAM for image, 24GB for video.
GitHub Copilot Agent Mode Exits Beta
GitHub Copilot's Agent Mode has officially exited beta, enabling developers to delegate multi-step coding tasks entirely to AI. The agent can plan implementations, create files, write tests, and submit pull requests. Available to all Copilot Individual and Business subscribers.
OpenAI Partners with Major Healthcare Systems for AI Diagnostics
OpenAI has announced partnerships with Mayo Clinic, Cleveland Clinic, and Johns Hopkins to deploy GPT-5 for medical image analysis and clinical decision support. Early trials show the AI matches specialist accuracy in radiology and pathology screening.
AI-Generated Content Now 15% of Published Scientific Papers
A comprehensive study published in Nature reveals that AI-assisted content now appears in approximately 15% of published scientific papers across major journals. Researchers use AI for literature reviews, data analysis, and manuscript drafting, raising both efficiency and integrity questions.
Suno V5 Introduces Multi-Track Editing and Stem Control
AI music platform Suno has released V5, introducing multi-track editing that lets users modify individual instruments, vocals, and effects independently. The update also includes real-time collaboration features and improved genre accuracy for 50+ music styles.
NVIDIA Reports Record Revenue Driven by AI Chip Demand
NVIDIA reported Q4 2026 revenue of $42 billion, up 65% year-over-year, driven by insatiable demand for H200 and B200 AI GPUs. Data center revenue alone exceeded $35 billion. CEO Jensen Huang projects continued growth as sovereign AI initiatives and enterprise adoption accelerate.
Apple Intelligence 2.0 Adds On-Device GPT-5 Mini
Apple has announced Apple Intelligence 2.0, featuring an on-device version of GPT-5 Mini that runs entirely on iPhone 16 and M4 Mac hardware. The update brings advanced writing tools, image generation, and a dramatically improved Siri with full conversational capabilities.
Mistral Releases Mixtral 3 with 128 Expert Architecture
French AI company Mistral has released Mixtral 3, featuring a novel 128-expert Mixture of Experts architecture. The model achieves Claude 4 Sonnet-level performance while maintaining competitive API pricing, positioning itself as the leading European AI alternative.
US Senate Passes Bipartisan AI Safety Framework
The US Senate has passed the AI Safety and Innovation Act with bipartisan support, establishing federal guidelines for AI deployment in critical infrastructure, healthcare, and finance. The legislation creates an AI Safety Board and mandates risk assessments for foundation models above a compute threshold.
ElevenLabs Launches Universal Voice Translation
ElevenLabs has launched a universal voice translation service that translates spoken content into 40+ languages while preserving the original speaker's voice, tone, and emotion. The service targets content creators, enterprises, and media companies for dubbing and localization.
AI Startups Raise Record $25B in Q1 2026
Global AI startup funding reached $25 billion in Q1 2026, surpassing the previous quarterly record by 40%. Infrastructure and tooling startups received the largest share (35%), followed by enterprise AI applications (28%) and AI safety/alignment (12%). Notable rounds include Anthropic, Cohere, and Scale AI.
Google Introduces AI Agents Marketplace in Workspace
Google has launched an AI Agents Marketplace within Google Workspace, allowing organizations to deploy pre-built AI agents for HR, IT support, sales, and customer service. Custom agents can be built using Gemini and deployed across Gmail, Docs, and Chat.
Report: AI Creates 2.3M New Jobs While Transforming 40% of Roles
A comprehensive World Economic Forum report finds that AI has created 2.3 million new jobs globally in 2025 while significantly transforming 40% of existing roles, with the greatest impact in knowledge work.
Cursor AI Reaches 5 Million Active Developers
AI-native code editor Cursor has surpassed 5 million monthly active developers, making it the fastest-growing developer tool in history. The company reports that Cursor users write 2.5x more code per day on average, with adoption strongest among TypeScript, Python, and Rust developers.
xAI Launches Grok 3.5 with Extended Thinking
Elon Musk's xAI has released Grok 3.5, featuring an extended thinking mode that shows step-by-step reasoning before answering complex questions. The model integrates real-time X data and shows improved performance in coding, math, and scientific reasoning.
China Unveils National AI Compute Infrastructure Plan
China has announced a $50 billion national AI compute infrastructure plan to build 10 government-backed AI computing centers across major cities. The initiative aims to provide subsidized AI compute access to Chinese companies and researchers, responding to US chip export restrictions.
Runway Launches Gen-4 Video Generation Model
Runway releases Gen-4, featuring multi-shot video generation, advanced camera controls, and character consistency across scenes, positioning it as a professional filmmaking tool.
Midjourney v7 Launches with Photorealistic Video Mode
Midjourney has released version 7, its most advanced model yet, featuring photorealistic image quality that is virtually indistinguishable from photographs. The update also introduces a video mode for generating short clips and improved text rendering within images.
Sam Altman Predicts AGI Within 3 Years
OpenAI CEO Sam Altman stated in a company blog post that he believes AGI (Artificial General Intelligence) could be achieved within three years, citing rapid progress in reasoning, multimodal understanding, and agent capabilities. The prediction has sparked debate among AI researchers about timelines and definitions.
Midjourney Launches v7 with Photorealistic Quality
Midjourney releases version 7 of its image generation model, achieving near-photorealistic quality with improved text rendering, better hands, and a new personalization feature that learns user style preferences.
Anthropic MCP Standard Adopted by 50+ AI Companies
Anthropic's Model Context Protocol (MCP) has been adopted by over 50 AI companies including Microsoft, Google, and Cursor, establishing it as the de facto standard for connecting AI models to external tools and data sources. The open standard now has 1,000+ community-built servers.
AI Energy Consumption Projected to Equal Small Country by 2027
A new report from the International Energy Agency estimates that global AI compute will consume 100+ TWh of electricity by 2027, equivalent to the energy consumption of the Netherlands. Major AI companies are investing in nuclear and renewable energy to power data centers sustainably.
NVIDIA Ships Blackwell Ultra GPUs for AI Training
NVIDIA begins shipping its Blackwell Ultra B300 GPUs, delivering 4x the AI training performance of the previous generation and addressing the massive compute demand from AI companies worldwide.
Perplexity AI Launches Enterprise Plan with SOC 2 Compliance
AI search engine Perplexity has launched its Enterprise plan featuring SOC 2 Type II compliance, SSO, admin controls, and internal knowledge base search. The plan costs $40/user/month and targets companies looking to replace traditional search with AI-powered research tools.
First AI-Discovered Drug Enters Phase III Clinical Trials
Insilico Medicine's AI-discovered drug candidate for idiopathic pulmonary fibrosis has entered Phase III clinical trials, marking a milestone for AI in drug discovery. The molecule was designed entirely by AI, reducing the typical discovery timeline from 5 years to 18 months.
Google Announces Imagen 4 with Industry-Leading Image Quality
Google unveils Imagen 4, its latest text-to-image model featuring photorealistic generation, superior text rendering, and advanced editing capabilities integrated directly into Google Workspace.
Google Previews Gemini 3 Pro with Next-Gen Capabilities
Google DeepMind unveils Gemini 3 Pro in preview, showcasing breakthrough performance in coding, mathematical reasoning, and multimodal understanding that sets new benchmarks across industry-standard evaluations.
EU AI Act Enforcement Begins with First Compliance Deadlines
The European Union begins enforcing key provisions of the AI Act, requiring companies to classify AI systems by risk level and implement transparency measures for high-risk applications.
OpenAI Releases Sora 2 for Professional Video Generation
OpenAI launches Sora 2, a major upgrade to its video generation model supporting up to 2-minute clips at 4K resolution with improved temporal consistency and professional-grade quality.
Anthropic Raises $5B in Series D Funding
Anthropic secures $5 billion in its latest funding round, valuing the AI safety company at over $60 billion. The funds will accelerate research into safe and interpretable AI systems.
OpenAI Releases Whisper v4 with 99% Accuracy
OpenAI releases Whisper v4, achieving near-human transcription accuracy across 100+ languages with real-time processing capability, making it the gold standard for speech-to-text applications.
Anthropic Releases Claude Opus 4 with Superior Safety
Anthropic launches Claude Opus 4, emphasizing constitutional AI principles with improved safety guardrails while delivering state-of-the-art performance in complex reasoning and long-document analysis.
US Updates AI Executive Order with New Safety Requirements
The Biden administration updates its AI executive order, introducing mandatory safety testing for frontier models above certain capability thresholds and requiring AI companies to share safety test results.
OpenAI Launches GPT-5 with Enhanced Reasoning
OpenAI has released GPT-5, its most capable model to date, featuring significantly improved reasoning, longer context windows up to 1M tokens, and native multimodal capabilities including image and audio understanding.
Meta Releases Llama 4 as Open-Source
Meta releases Llama 4, its latest open-source large language model, available for commercial use. The model matches proprietary competitors on key benchmarks while remaining freely accessible to developers worldwide.
DeepSeek V3 Released with Open Weights
Chinese AI lab DeepSeek releases V3 with fully open weights, demonstrating performance competitive with GPT-5 on coding and reasoning benchmarks at a fraction of the training cost.