Gemini 2.0 is now available to everyone
Google DeepMind announced new updates to Gemini 2.0 Flash and introduced Gemini 2.0 Flash-Lite and Gemini 2.0 Pro Experimental.
Google DeepMind·5 Feb
Use this view to inspect the underlying evidence corpus. For ranked developments, decision posture and interpretation, use Signals.
Raw feed or Signals?
Raw feed is chronological evidence. Signals ranks and interprets material change.
Google DeepMind announced new updates to Gemini 2.0 Flash and introduced Gemini 2.0 Flash-Lite and Gemini 2.0 Pro Experimental.
OpenAI partnered with the California State University (CSU) system to deploy ChatGPT access for 500,000 students and faculty across 23 campuses.
OpenAI's Frontier Lab is collaborating with Bain & Company to apply advanced AI research to analyze complex industry trends.
OpenAI terminated accounts originating from China that were using its models to plan and optimize surveillance activities.
OpenAI terminated accounts linked to a deceptive employment scheme associated with North Korean IT worker activity.
OpenAI terminated Cambodian accounts using its models to translate and generate scripts for pig-butchering financial scams.
OpenAI terminated accounts linked to state-sponsored DPRK threat actors using LLMs for malware research, phishing, and cyber reconnaissance.
The EU AI Act's 2025 AI Action Summit in Paris, 10-11 February, will establish implementation deliverables for the regulation.
OpenAI released a system card for its o3-mini model, detailing safety evaluations, external red teaming, and Preparedness Framework assessments.
METR Research's preliminary evaluations of Anthropic's Claude 3.5 Sonnet and OpenAI's o1 found no significant evidence of dangerous capabilities.
Mistral AI quietly released 'Mistral Small 3', a new foundational model, with early benchmarks suggesting strong performance for its size.
OpenAI partners with U.S. National Laboratories to apply frontier reasoning models for scientific research, aiming for breakthroughs.
OpenAI introduced 'ChatGPT Gov', a version of ChatGPT tailored for government use with enhanced security and data handling features.
China's DeepSeek AI claims to have developed high-performing models efficiently without advanced chips, potentially lowering compute costs.
DeepSeek, a Chinese AI model, is gaining attention for strong performance despite using less advanced chips, leading to industry praise.
Meta plans to spend $60-$65 billion on AI and data centers, signaling accelerating tech industry investments in artificial intelligence.
OpenAI published an "Operator System Card" detailing its multi-layered safety framework, including mitigations for prompt engineering and privacy.
SK Hynix, an Nvidia supplier, posted record quarterly and annual profits in Q4 2024 driven by robust demand for AI-related memory chips.
WSJ's latest Market Talk covers technology, media, and telecom, including insights on Netflix, US advertising spending, and the AI joint venture Stargate.
Bertelsmann, a global media and services firm, announced broad integration of OpenAI technology across its diverse brands for creativity and productivity.
OpenAI, Oracle, and SoftBank pledged up to $500 billion in AI infrastructure investments in the U.S. at a White House event.
Hugging Face partnered with FriendliAI to offer optimized inference for open-source models directly on the Hugging Face Hub.
OpenAI published a note on 'Stargate Infrastructure,' signaling intent to build massive data centers for AGI and seeking industry partners.
TikTok service will be restored for U.S. users following an executive order from the president-elect, with a long-term solution planned.
TikTok told users its app is unavailable; former President Trump indicated he would likely intervene regarding the platform.
OpenAI highlighted the potential of personalized AI, allowing models to adapt to individual user preferences and data for improved utility.
METR Research highlights that frontier AI safety policies focus on pre-deployment testing for public models, a standard across many products.
Chip Huyen outlines common pitfalls in generative AI application development, including misapplying GenAI and challenges in evaluation.
Hugging Face Text Generation Inference now supports multiple backends (TRT-LLM, vLLM) for improved performance and flexibility.
OpenAI partnered with Axios and other publishers to integrate news content into ChatGPT and support AI adoption in newsrooms via grants and tools.
Hugging Face claims new techniques accelerate static embedding model training by 400x using Sentence Transformers.
Adebayo Ogunlesi, Chairman and CEO of Global Infrastructure Partners, has joined OpenAI’s Board of Directors as an independent director.
Hugging Face blog discusses the rise of AI agents, their current capabilities, and future implications for enterprise applications.
An exploration of AI-powered reading features, including summarization, interactive Q&A, and content organization, for improved knowledge consumption.
Hugging Face analysis links increased LLM training CO₂ emissions with marginal performance gains for larger models, suggesting diminishing returns.
New research from Future of Life validates long-standing concerns regarding the difficulty of safely constraining powerful AI systems, raising control questions.
OpenAI announced a new governance structure, intending to strengthen its non-profit mission through for-profit success and a new 'safety and security committee.'
NVIDIA released LogitsProcessorZoo on Hugging Face, offering advanced control over language model output generation through custom logits processors.
OpenAI introduces "deliberative alignment" for o1 models, teaching safety specifications and reasoning for enhanced safety.
Dr. Rumman Chowdhury discussed algorithmic auditing and 'right to repair' AI systems at an EU AI Act 'Safety Breakfast' event.
The 'AI Snake Oil' newsletter argues that recent AI progress is showing diminishing returns, questioning the rapid advancement narrative.
Google DeepMind introduces FACTS Grounding, a new benchmark and leaderboard to evaluate LLM factuality and hallucination against source material.
Hugging Face benchmarked language model inference performance on Intel 5th Gen Xeon processors on Google Cloud Platform.
OpenAI announced o1, a new model, alongside Realtime API improvements and a new fine-tuning method for developers.
Analysis of 78 election deepfakes found most were low-tech fakes, suggesting misinformation is a human problem, not primarily an AI problem.
OpenAI claims Elon Musk himself proposed a for-profit structure in 2017, contradicting his current lawsuit and public statements.
Google DeepMind announced Gemini 2.0, a new multimodal AI model, claiming increased capabilities for agentic applications.
Future of Life Institute released an AI Safety Index evaluating leading AI companies; findings indicate most neglect safety concerns.
Zalando claims to enhance its customer retail experience by powering its Assistant with OpenAI's GPT-4o mini.
Hugging Face is making its open-source models available through Amazon Bedrock, allowing enterprise access to OSS models via a managed AWS service.
© 2026 OneBench: AI Insights. All rights reserved.
Evidence before opinion