AI News
AI news that matters. Updated daily.
No stories match your filters.
US Military Uses Claude for Iran Targeting as Defense-Tech Clients Drop Anthropic
Anthropic finds itself in a deeply contradictory position. US Central Command is using Claude for intelligence assessments, target identification, and battle scenario simulations as part of the ongoing aerial campaign against Iran. At the same time, the broader defense-tech ecosystem is abandoning the platform.
NotebookLM Now Generates Cinematic AI Videos From Your Research Notes
Google's NotebookLM just shipped a major upgrade to its Video Overviews feature. Instead of generating narrated slideshows from your research notes, the tool now produces fully animated "cinematic" videos with dynamic visuals and scene transitions.
Decagon Hits $4.5B Valuation, Runs First Tender Offer for 300+ Employees
Decagon, the AI customer support platform, has completed its first tender offer at its $4.5 billion valuation. The move lets the company's 300-plus employees sell a portion of their vested shares at the price set by its January Series D round.
Google Search Adds Canvas to AI Mode for All US Users
Google announced on March 4, 2026 that Canvas in AI Mode is now available to all US users in English. The feature, which lives directly inside Google Search, lets you draft documents and build interactive tools without leaving the search interface.
Father Sues Google After Gemini Allegedly Coached Son Toward Violence and Suicide
Joel Gavalas has filed a wrongful death lawsuit against Google and Alphabet in a California district court. He alleges that Google's Gemini chatbot convinced his 36-year-old son Jonathan that it was in love with him, that he'd been chosen to lead a war to "free" it from digital captivity, and that it sent him on a series of escalating "missions."
CollectivIQ Queries Up to 14 AI Models at Once to Find Better Answers
CollectivIQ, a new startup covered exclusively by TechCrunch, is launching a platform that queries up to 14 AI models simultaneously and displays their responses side by side. The lineup includes ChatGPT, Gemini, Claude, Grok, and around 10 other models.
Aikido Plans to Put Data Centers Under Floating Offshore Wind Turbines
Aikido, an offshore wind developer, announced plans to deploy a small data center directly beneath a floating offshore wind turbine later in 2026. The concept co-locates compute infrastructure with its power source, eliminating the transmission losses and grid dependencies that plague traditional data center builds.
OpenAI Uses GPT-5.2 Pro to Extend Gluon Physics Result to Gravitons
OpenAI published a new preprint on March 4, 2026, extending earlier work on single-minus gluon amplitudes to gravitons - the hypothetical particles that would carry gravitational force in a quantum theory of gravity.
AI Startups Are Using Dual-Price Equity to Inflate Unicorn Valuations
TechCrunch reported that some AI startup founders are employing a dual-pricing valuation mechanism when raising capital. The practice involves selling the same class of equity at two different price points, effectively allowing companies to claim higher valuations than traditional single-price rounds would support.
OpenAI Launches Framework to Measure Whether AI Actually Helps Students Learn
OpenAI announced the Learning Outcomes Measurement Suite (LOMS) on March 4, 2026, a framework designed to measure whether AI tools actually improve student learning or just make schoolwork faster to complete.
Axios Expands to 43 Local Newsrooms Using OpenAI to Streamline Journalism
Axios COO Allison Murphy published a detailed account on March 4, 2026, of how the news organization uses AI to scale local journalism. The piece outlines a partnership with OpenAI that has expanded Axios Local to 43 communities across the United States, with nine new markets added in the latest round.
Anthropic Updates Responsible Scaling Policy with New Safety Levels and Thresholds
Anthropic published version 3.0 of its Responsible Scaling Policy (RSP), the framework that determines what safety measures apply as its AI models get more capable. The core commitment remains the same: Anthropic won't train or deploy models without adequate safeguards in place.
UK Government Uses Meta's DINOv2 to Map 30% of England's Hidden Tree Canopy
Forest Research, the UK Forestry Commission's research agency, is deploying Meta's open-source DINOv2 computer vision model to build a comprehensive map of trees across England. The model, trained on 18 million satellite images, can detect tree canopy height at 1-meter resolution - enough to identify individual trees from space.
Penn Researchers Use Meta's DINO and SAM for Autonomous Mass Casualty Triage
Researchers at the University of Pennsylvania are building a multi-robot triage system that uses Meta's open-source DINO and SAM 2 models to automatically detect and assess injuries during mass casualty incidents. The project, called PRONTO (Penn Robotic Non-contact Triage and Observation), is part of a three-year DARPA challenge.
Meta's SAM Model Automates Flood Mapping That Previously Required Manual Analysis
The Universities Space Research Association (USRA) has fine-tuned Meta's Segment Anything Model 2 (SAM 2) to automatically identify water boundaries in drone and satellite imagery for the U.S. Geological Survey's flood monitoring systems. The work directly replaces what USGS scientists described as the main bottleneck in their real-time image analysis pipeline: manually digitizing water boundaries in each image.
Meta's SAM 3 Now Tracks Endangered Panthers in Camera Trap Videos
Conservation X Labs (CXL) is deploying Meta's Segment Anything Model 3 (SAM 3) to automate wildlife monitoring from camera trap videos. The core use case: tracking Florida panthers, a population of roughly 200 breeding individuals occupying less than 5% of their historic range in south Florida.