Expanding Stargate to Michigan
OpenAI expands Stargate project to Michigan with a new one-gigawatt campus. The infrastructure strengthens US AI compute capacity and will create jobs in the Midwest.
88 articles
OpenAI expands Stargate project to Michigan with a new one-gigawatt campus. The infrastructure strengthens US AI compute capacity and will create jobs in the Midwest.
OpenAI introduces Aardvark, an autonomous security researcher that detects, validates, and fixes software vulnerabilities at scale. The system is in private beta with sign-up for early testing.
Hugging Face analyzes agent generalization in MiniMax M2, questioning current alignment metrics. The study examines how agents perform beyond training data and proposes more robust evaluation criteria to measure true generalization.
OpenAI introduces OWL, the architecture powering Atlas, its ChatGPT-based browser. The system decouples Chromium to enable faster startup, richer UI, and agentic browsing capabilities integrated with ChatGPT.
Google DeepMind launches the AI for Math Initiative, bringing together prestigious research institutions to explore AI's role in mathematical research and accelerate scientific discovery.
Hugging Face analyzes shifts in the global AI compute landscape. The article examines trends in infrastructure, resource allocation, and implications for language model accessibility.
OpenAI releases two open-weight reasoning models: gpt-oss-safeguard-120b and gpt-oss-safeguard-20b, post-trained to label content according to provided policies. Technical report covers capabilities and baseline safety evaluations against underlying gpt-oss models.
OpenAI releases gpt-oss-safeguard, open-weight reasoning models for safety classification enabling developers to apply and iterate on custom policies. Developers gain control over safety rules without relying on proprietary APIs.
Hugging Face and NVIDIA demonstrate an end-to-end pipeline for healthcare robot development: virtual environment simulation, reinforcement learning training, and physical deployment. Uses Isaac Sim and AI models to automate medical tasks.
NVIDIA Isaac for Healthcare enables building healthcare robots in simulation before real-world deployment. The platform provides modeling, physics, and AI tools to develop autonomous robotic systems for medical environments.
DNP rolls out ChatGPT Enterprise across ten core departments: patent research 95% faster, 10x processing volume, 87% automation, 70% knowledge reuse achieved in three months.
IBM and Hugging Face release Granite 4.0 Nano, an ultra-lightweight model optimized for on-device inference. The model drastically reduces size while maintaining reasoning and code generation capabilities.
Doppel, an AI defense system using GPT-5 and reinforcement fine-tuning, detects and blocks deepfake and impersonation attacks before spread. Reduces analyst workload by 80% and response times from hours to minutes.
OpenAI announces recapitalization to strengthen mission-focused governance and expand resources ensuring AI benefits everyone while advancing responsible innovation. No financial details or timeline provided in excerpt.
Microsoft and OpenAI sign a new agreement strengthening their long-term partnership. The deal expands joint innovation and reaffirms commitment to responsible AI progress, with no specific financial or technical details disclosed.
Hugging Face releases guidelines for ethical voice cloning with explicit consent. The platform introduces verification mechanisms and tools to ensure voice synthesis models respect original speakers' rights.
OpenAI submits strategic investment plan to U.S. White House on energy and infrastructure expansion needed to sustain American AI leadership and economic growth. Document outlines capacity and workforce readiness requirements.
OpenAI partnered with 170+ mental health experts to enhance ChatGPT's distress recognition and empathetic responses. Unsafe responses reduced by up to 80%, with improved guidance toward real-world support resources.
OpenAI releases addendum to GPT-5 system card documenting improvements in handling sensitive conversations: new benchmarks for emotional reliance, mental health, and jailbreak resistance.
Hugging Face releases huggingface_hub v1.0 after five years of development. The Python library has become the core infrastructure for sharing and accessing open-source models, datasets, and spaces. This release marks API stability and solidifies Hugging Face's role as a major collaborative platform for open-source ML.
Hugging Face introduces dataset streaming, reducing memory consumption by 100x. Data loads on-demand without prior download, accelerating training and experimentation on large corpora.
Steuerrecht.com integrates ChatGPT Business to automate tax research and accelerate delivery of client-ready legal analysis for law firms.
Google DeepMind introduces T5Gemma, a new collection of encoder-decoder Gemma models. These LLMs combine T5 architecture with Gemma foundation for translation, summarization, and text generation tasks.
Google DeepMind announces MedGemma, a collection of multimodal open-source models specialized for health AI development. These represent DeepMind's most capable open models for medical application development.
Google DeepMind releases Gemma 3n with a developer guide. The model is designed based on feedback from the developer community that shaped Gemma's evolution.
Gemini 2.5 Flash-Lite exits preview and becomes generally available for production use. This lightweight model delivers high quality at low cost, featuring a 1 million-token context window and multimodality.
Google DeepMind partnered with director Darren Aronofsky and Eliza McNitt to produce the film 'ANCESTRA' combining the Veo video model with live-action filmmaking. A team of over 200 people explored integrating generative AI into film production.
Google DeepMind releases AlphaEarth Foundations, an AI model integrating petabytes of Earth observation data to generate a unified data representation. This tool revolutionizes global mapping and monitoring at unprecedented scale.
Google DeepMind launches Backstory, an experimental AI tool enabling exploration of context and origin of images found online. The tool helps understand where images come from and their usage context.
Gemini with Deep Think reaches gold-medal standard at the International Mathematical Olympiad. The model solves all six exceptionally difficult problems from the world's most prestigious annual mathematics competition.
Google DeepMind introduces Aeneas, an AI model specialized in contextualizing ancient inscriptions. The tool helps historians interpret, attribute, and restore fragmentary texts.
Google DeepMind introduces Genie 3, a world model that generates dynamic, navigable environments in real time at 24 fps in 720p resolution, maintaining consistency for several minutes.
Google DeepMind releases Perch, an AI model for bioacoustics audio analysis. The tool speeds up identification of endangered species (Hawaiian honeycreepers, coral reefs) for conservationists.
Google DeepMind explores using AI to enhance perception and analysis of the universe. The article details how machine learning models can process complex cosmological data and reveal hidden structures in astronomical observations.
Gemini 2.5 Deep Think achieves gold-medal level performance at the International Collegiate Programming Contest World Finals, demonstrating a major breakthrough in abstract problem-solving capabilities.
Google DeepMind introduces an AI method to solve century-old problems in fluid dynamics. The approach could help mathematicians and physicists tackle long-standing challenges in mathematics, physics and engineering.
LeRobot v0.4.0 enhances open-source robot learning with improved modeling and training capabilities. Hugging Face's platform integrates pre-trained models and datasets to accelerate autonomous robot development.
Google DeepMind strengthens its Frontier Safety Framework to identify and mitigate severe risks from advanced AI models. Updated safety strategy for frontier systems.
Google DeepMind releases Gemini Robotics 1.5, an AI model for robots capable of perceiving, planning, reasoning, using tools and acting. The system targets complex multi-step tasks in the physical world.
Google DeepMind introduces CodeMender, an AI agent that automatically detects and fixes critical software vulnerabilities. The tool uses advanced AI techniques to identify security flaws and propose patches.
Google DeepMind partners with Commonwealth Fusion Systems to accelerate fusion energy development. The partnership applies AI to technical challenges in civilian nuclear fusion.
Google DeepMind rolls out Deep Think in the Gemini app for Google AI Ultra subscribers. Select mathematicians gain access to the full Gemini 2.5 Deep Think model entered in the IMO competition.
Google DeepMind releases Game Arena, an open-source platform for rigorous evaluation of AI models through head-to-head comparisons in environments with clear winning conditions.
Google DeepMind releases Gemma 3 270M, a compact 270-million parameter model designed for hyper-efficient AI. This addition expands the Gemma 3 toolkit with a lightweight option optimized for resource-constrained deployments.
Google DeepMind upgrades native image editing capabilities in the Gemini app. Users can now transform images in new ways directly within the interface.
Google DeepMind introduces VaultGemma, a language model trained from scratch with differential privacy. It is the most capable model ever created under this privacy constraint.
Google DeepMind introduces Gemini 2.5 Computer Use, a specialized model built on Gemini 2.5 Pro capabilities to power agents that interact with user interfaces. Available in preview via API.
Google DeepMind rolls out Veo 3.1 with advanced creative capabilities providing users greater control over video generation.
Google DeepMind launches a 27 billion parameter Gemma model specialized for single-cell analysis. The model helped identify a new potential cancer therapy pathway.
OpenAI acquires Software Applications Incorporated, maker of Sky—a natural language interface for macOS. Sky's deep macOS capabilities will be integrated into ChatGPT to make AI more intuitive, contextual, and action-oriented on desktop.
Consensus integrates GPT-5 and OpenAI's Responses API to power a multi-agent research assistant that reads, analyzes, and synthesizes scientific evidence in minutes. The platform serves 8 million researchers to accelerate scientific discovery.
Hugging Face introduces OpenEnv, an open platform for building and testing AI agents. The initiative aims to create a collaborative ecosystem of standardized benchmarks and environments to evaluate autonomous agent performance.
OpenAI launches Company Knowledge for ChatGPT Business, Enterprise, and Edu tiers: integrates business data with citations, access controls, and compliance. Enables contextual answers from internal company apps.
OpenAI releases an economic blueprint for South Korea outlining sovereign AI capabilities and strategic partnerships to drive growth. The plan focuses on scaling trusted AI adoption across the country. No specific figures or model details provided in the excerpt.
OpenAI expands UK partnership with Ministry of Justice agreement to deploy ChatGPT to civil servants. Introduces UK data residency for ChatGPT Enterprise, ChatGPT Edu, and API Platform to support sovereign AI adoption and data security.
OpenAI releases an economic blueprint for Japan outlining how to leverage AI for innovation and competitiveness. The plan addresses sustainable, inclusive growth through AI integration in the Japanese economy. No specific metrics or technical details disclosed in the excerpt.
Hugging Face and VirusTotal partner to strengthen AI security. The collaboration integrates VirusTotal's malware detection capabilities into the Hugging Face ecosystem, enabling automated scanning of AI artifacts against known threats.
Sentence Transformers, the open-source library for generating text embeddings, officially joins Hugging Face. The development team integrates the platform to accelerate innovation in semantic representations and improve accessibility of embedding models.
OpenAI discontinues ChatGPT on WhatsApp after January 15, 2026. Users must link their ChatGPT account to continue conversations across other devices.
OpenAI launches ChatGPT Atlas, a browser with built-in ChatGPT for instant answers and web page summaries. Available now on macOS with user-controlled privacy settings.
Hugging Face releases AI Sheets, a tool embedding AI vision capabilities into spreadsheets. Enables image analysis, classification, and data extraction directly in Excel/Sheets without coding.
Hugging Face showcases open-source models to enhance OCR pipelines. The article details integration of vision and text processing models to extract and structure document content with accuracy.
Datalab releases two new models for text extraction from documents and images: conversion to markdown or line-level polygon extraction.
Hugging Face presents an AI application to detect food allergens. The system uses computer vision to analyze food images and identify potential allergens, improving food safety for people with allergies.
Google Cloud C4 delivers 70% TCO improvement for GPT open-source models in partnership with Intel and Hugging Face. The optimization combines Google Cloud's C4 infrastructure with Intel processors to significantly reduce inference and deployment costs.
Google Veo 3.1 adds reference image control, first/last frame control, and enhanced image-to-video capabilities. Complete guide to new features.
Plex Coffee deploys ChatGPT Business to centralize operational knowledge, speed up staff training, and preserve service quality during expansion. Retail/hospitality use case demonstrating LLM integration into business workflows without quantified metrics.
Hugging Face publishes a guide to run vision-language models (VLMs) on Intel CPUs in three steps. The method optimizes performance on CPU without dedicated GPU.
OpenAI launches Expert Council on Well-Being and AI with leading psychologists and clinicians to guide ChatGPT's role in emotional health support, particularly for teens. The council shapes safer, more caring AI experiences.
OpenAI and Sur Energy are exploring Argentina's first Stargate project to position the country as a Latin American leader in AI, sustainable infrastructure, and digital innovation. No technical details, timeline, or investment figures disclosed.
Nvidia releases Nemotron-Personas-India, a synthetic dataset for training sovereign AI models in India. The dataset contains personas and generated data reflecting Indian cultural and linguistic context, enabling local organizations to develop tailored models without external infrastructure dependency.
OpenAI and Broadcom announce multi-year partnership to deploy 10 gigawatts of OpenAI-designed AI accelerators by 2029. The companies will co-develop next-generation systems and Ethernet solutions for scalable, energy-efficient AI infrastructure.
HYGH accelerates software development and campaign delivery using ChatGPT Business, reducing turnaround times and scaling output. No specific metrics or technical details provided in excerpt.
OpenAI releases methodology to evaluate political bias in ChatGPT through real-world testing methods. The approach aims to improve objectivity and reduce systematic model bias.
HiBob deployed 2,500 custom GPTs using ChatGPT Enterprise to scale internal AI adoption, optimize HR workflows, and integrate AI features into the Bob platform. Use case: product scaling and team growth without disclosed ROI metrics.
Hugging Face introduces BigCodeArena, a code generation evaluation platform based on actual code execution. It measures end-to-end performance rather than textual comparison, enabling objective assessment of generation quality.
OpenAI releases October 2025 report on detecting and disrupting malicious AI use. Details enforcement actions, policy implementation, and user protection measures against real-world harms.
OpenAI Codex now generally available with Slack integration, dedicated SDK, and admin tools (usage dashboards, workspace management) to simplify large-scale deployment.
OpenAI launches Apps, a new category of conversational applications embedded directly in ChatGPT. A public SDK in preview enables developers to start building immediately.
AMD and OpenAI announce multi-year partnership to deploy 6 gigawatts of Instinct GPUs, starting with 1 GW in 2026, to power OpenAI's next-generation AI infrastructure and accelerate global AI innovation.
OpenAI releases AgentKit, a toolkit to speed prototype-to-production development. Includes expanded evals capabilities and reinforcement fine-tuning (RFT) for agents. Official announcement with limited technical details in excerpt.
OpenAI and Allied for Startups release the Hacktivate AI report featuring 20 actionable policy recommendations to accelerate AI adoption in Europe, boost competitiveness, and empower innovators.
Wrtn reaches 6.5M users in Korea using GPT-5, building "Lifestyle AI" combining productivity, creativity, and learning. The company is expanding across East Asia.
IBM releases Granite 4.0, now available on Replicate. The model provides enhanced capabilities for generative AI tasks. Direct integration via Replicate platform for streamlined access and deployment.
OpenAI partners with Japan's Digital Agency to deploy generative AI in public services, strengthen international AI governance, and promote safe, trustworthy AI adoption globally.
Hugging Face releases dots.ocr, a state-of-the-art OCR model optimized for Core ML. The model runs on-device on iOS/macOS with high accuracy and low latency, without server dependency.
Samsung and SK Telecom join OpenAI's Stargate initiative to expand global AI infrastructure. Partners will scale advanced memory chip production and build next-generation data centers in South Korea.
Hugging Face introduces RTEB, a new standard for evaluating retrieval systems. RTEB provides standardized and reproducible benchmarks to measure the performance of embedding models and RAG pipelines.