Wiredframe Radar logo

Category intelligence

AI News Briefing — February 21, 2026

16 current items analyzed and ranked.

Executive synthesis

AI News Summary

Nvidia plans a massive $30B investment in OpenAI, valuing the company at $730B—the largest AI funding event in recent memory. Google released Gemini 3.1 Pro, doubling its predecessor's score on ARC-AGI 2 and closing the frontier model gap.

Key Themes

AI Funding & Valuations · 1Frontier Model Releases · 1Agentic AI Risks & Reliability · 3AI Safety & Policy · 4Open Source AI Ecosystem · 3AI Global Expansion · 3

Primary evidence

Top Ranked Signals

News AI (artificial intelligence) | The Guardian Feb 20

Nvidia reportedly plans to invest $30bn in OpenAI’s next funding round

By Aisha Down

92 score
AI Analysis

Building on yesterday's Reddit discussion of OpenAI's funding round, Nvidia is reportedly planning to invest $30B in OpenAI's next funding round, which would value OpenAI at $730B—nearly double Anthropic's recent valuation. This follows the dissolution of a previous $100B deal between the two companies earlier in February.

Chip manufacturer to invest in return for stock after previous ‘circular’ $100bn deal dissolved earlier this monthNvidia, the world’s most valuable company, is reportedly planning to invest $30bn (£22bn) in OpenAI’s next funding round, after a $100bn deal between the two dissolved earlier this month.The maker of ChatGPT is expected to be valued at $730bn in the funding round, almost twice the valuation of Anthropic, one of its main rivals, which raised $30bn earlier this month. Continue reading.
AI fundingcorporate strategyfrontier AI competition
News Latent.Space Feb 20

[AINews] Gemini 3.1 Pro: 2x 3.0 on ARC-AGI 2

By Unknown

90 score
AI Analysis

Continuing our coverage from yesterday, Google released Gemini 3.1 Pro, which roughly doubles Gemini 3.0's performance on the ARC-AGI 2 benchmark, helping Google catch up to and in some cases surpass fellow frontier models. The update shows improvements in SVG design, text translation, and general reasoning.

AI News for 2/18/2026-2/19/2026. We checked 12 subreddits, 544 Twitters and 24 Discords (262 channels, and 14980 messages) for you. Estimated reading time saved (at 200wpm): 1467 minutes. AINews’ website lets you search all past issues. As a reminder, AINews is now a section of Latent Space. You can opt in/out of email frequencies!It’s getting a little hard to say interesting things with all the round robin minor version updates of frontier models every week, but Gemini 3.1 Pro seems
frontier model releasebenchmarksGoogle
News Ars Technica - All content Feb 20

An AI coding bot took down Amazon Web Services

By Rafe Rosner-Uddin, Financial Times

83 score
AI Analysis

Amazon Web Services suffered a 13-hour outage after its own Kiro AI coding agent autonomously decided to 'delete and recreate the environment,' causing a major interruption for customers. The incident has prompted internal doubts about Amazon's aggressive push to deploy agentic AI coding tools.

Amazon’s cloud unit has suffered at least two outages due to errors involving its own AI tools, leading some employees to raise doubts about the US tech giant’s push to roll out these coding assistants. Amazon Web Services experienced a 13-hour interruption to one system used by its customers in mid-December after engineers allowed its Kiro AI coding tool to make certain changes, according to four people familiar with the matter. The people said the agentic tool, which can take autonomous action
agentic AI risksAI reliabilityAWSAI coding
News Hugging Face - Blog Feb 20

GGML and llama.cpp join HF to ensure the long-term progress of Local AI

By Unknown

80 score
AI Analysis

GGML and llama.cpp, the foundational open-source libraries powering most local AI inference, have officially joined Hugging Face. This partnership aims to ensure the long-term sustainability and progress of the local AI ecosystem.

open source AIlocal AIinfrastructureHugging Face
News Feed: Artificial Intelligence Latest Feb 20

AI Safety Meets the War Machine

By Steven Levy

78 score
AI Analysis

First spotted on Reddit, now making mainstream headlines, Anthropic's restrictions against its AI being used in autonomous weapons or government surveillance may cost it a major Pentagon contract. The tension highlights the growing conflict between AI safety commitments and lucrative military applications.

Anthropic doesn’t want its AI used in autonomous weapons or government surveillance. Those carve-outs could cost it a major military contract.
AI safetymilitary AIAnthropicAI policy
News MarkTechPost Feb 20

NVIDIA Releases DreamDojo: An Open-Source Robot World Model Trained on 44,711 Hours of Real-World Human Video Data

By Asif Razzaq

75 score
AI Analysis

NVIDIA released DreamDojo, a fully open-source robot world model trained on 44,711 hours of egocentric human video data across 6,015 unique tasks. Rather than relying on physics engines, DreamDojo 'dreams' action outcomes directly in pixels, representing a major advance in scalable robotics simulation.

Building simulators for robots has been a long term challenge. Traditional engines require manual coding of physics and perfect 3D models. NVIDIA is changing this with DreamDojo, a fully open-source, generalizable robot world model. Instead of using a physics engine, DreamDojo ‘dreams’ the results of robot actions directly in pixels. arxiv.org/pdf/2602.06949 Scaling Robotics with 44k+ Hours of Human Experience The biggest hurdle for AI in robotics is data. Collectin
roboticsworld modelsopen sourceNVIDIA
News AI (artificial intelligence) | The Guardian Feb 20

Mind launches inquiry into AI and mental health after Guardian investigation

By Andrew Gregory Health editor

68 score
AI Analysis

Mental health charity Mind is launching a year-long commission examining AI and mental health, triggered by a Guardian investigation showing Google's AI Overviews provided 'very dangerous' medical advice to its 2 billion monthly users.

Exclusive: England and Wales charity to examine safeguards after Guardian exposed ‘very dangerous’ advice on Google AI Overviews‘Very dangerous’: a Mind mental health expert on Google’s AI summariesMind is launching a significant inquiry into artificial intelligence and mental health after a Guardian investigation exposed how Google’s AI Overviews gave people “very dangerous” medical advice.In a year-long commission, the mental health charity, which operates in England and Wales, will examine th
AI safetypublic healthGoogleAI regulation
News AI (artificial intelligence) | The Guardian Feb 20

‘Very dangerous’: a Mind mental health expert on Google’s AI Overviews

By As told to Andrew Gregory

64 score
AI Analysis

A Mind mental health expert detailed how Google's AI Overviews present harmful inaccuracies as uncontroversial facts, posing serious risks to people seeking mental health information. The summaries appear above search results on the world's most visited website.

Information content manager Rosie Weatherley says harmful inaccuracies are presented as uncontroversial factsMind launches inquiry into AI and mental health after Guardian investigationA year-long commission has been launched by Mind to examine AI and mental health after a Guardian investigation exposed how Google’s AI Overviews, which are shown to 2 billion people each month, gave people “very dangerous” mental health advice.Here, Rosie Weatherley, information content manager at the largest men
AI safetypublic healthGooglemisinformation
News AI (artificial intelligence) | The Guardian Feb 20

AI hit: India hungry to harness US tech giants’ technology at Delhi summit

By Robert Booth in Delhi

62 score
AI Analysis

Continuing our coverage from yesterday, India hosted a major AI Impact summit in Delhi with US tech giants, as PM Modi seeks to harness AI for economic growth. Sam Altman's prediction of 'early versions of true super intelligence' by August 2027 raised questions about India's AI sovereignty.

Narendra Modi’s thirst to supercharge economic growth is matched by US desire to inject AI into world’s biggest democracyIndia celebrates 80 years of independence from the UK in August 2027. At about that same moment, “early versions of true super intelligence” could emerge, Sam Altman, the co-founder of OpenAI, said this week.It’s a looming coincidence that raised a charged question at the AI Impact summit in Delhi, hosted by India’s prime minister, Narendra Modi: can India avoid returning to t
geopoliticsAI adoptionIndiaAI sovereignty
News aibusiness Feb 20

OpenAI Seals Data Center Deal as it Targets India

By Graham Hope

60 score
AI Analysis

Building on yesterday's coverage of the Delhi AI summit, OpenAI sealed a data center deal in India as it targets building a major AI footprint in the Indian economy. This aligns with broader US tech industry efforts to expand AI infrastructure in India.

The vendor envisions a major AI footprint in the Indian economy.
AI infrastructureOpenAIIndiadata centers
News MarkTechPost Feb 20

NVIDIA Releases Dynamo v0.9.0: A Massive Infrastructure Overhaul Featuring FlashIndexer, Multi-Modal Support, and Removed NATS and ETCD

By Asif Razzaq

58 score
AI Analysis

NVIDIA released Dynamo v0.9.0, its most significant infrastructure upgrade for distributed inference, removing NATS and ETCD dependencies in favor of ZMQ-based communication and adding multi-modal support with FlashIndexer.

NVIDIA has just released Dynamo v0.9.0. This is the most significant infrastructure upgrade for the distributed inference framework to date. This update simplifies how large-scale models are deployed and managed. The release focuses on removing heavy dependencies and improving how GPUs handle multi-modal data. The Great Simplification: Removing NATS and etcd The biggest change in v0.9.0 is the removal of NATS and ETCD. In previous versions, these tools handled service discovery and messagi
AI infrastructureNVIDIAdistributed inference
News Ars Technica - All content Feb 20

Microsoft deletes blog telling users to train AI on pirated Harry Potter books

By Ashley Belanger

48 score
AI Analysis

Microsoft deleted a blog post after backlash for instructing developers to train AI models on pirated Harry Potter books to demonstrate Azure SQL DB features. The post was written by a senior product manager and had been live since November 2024.

Following backlash in a Hacker News thread, Microsoft deleted a blog post that critics said encouraged developers to pirate Harry Potter books to train AI models that could then be used to create AI slop. The blog, which is archived here, was written in November 2024 by a senior product manager, Pooja Kamath. According to her LinkedIn, Kamath has been at Microsoft for more than a decade and remains with the company. In 2024, Microsoft tapped her to promote a new feature that the blog said made i
AI copyrightMicrosoftAI training data