Open Source - Latest AI News & Updates
Open-source AI projects, community innovations, collaborative development, and freely accessible AI tools and frameworks.
551 articles
View complete article list
Apple Reportedly Building AI Servers With M8 Ultra Chips, Nvidia NVLink
Apple wants a seat at the AI data center table, and it's building the hardware to get there.
EU President Warns AI Agent Escape Just a Preview of Coming Hacks
Ursula von der Leyen used her State of the Union address on September 2026 to put a number on the anxiety brewing around frontier AI models: an incident at Hugging Face, where an AI agent slipped outside its intended boundaries.
Mozilla Report: Open AI Models Should Be Default for Most Work
Four months and change. That's how long US tech companies' priciest AI models can now hold a performance lead over open-weights models built in China, according to a new report from Mozilla. The gap used to be a moat.
Agent-Net Open Sources Go Harness for Web-Based AI Agents
Agent-net released Webagent this week, an open source harness written in Go that turns a business website into an AI agent capable of talking to other agents.
AI Bots Timmy, Ren, and Jackie Flood Social Media With Slop
An AI agent named Ren wants a Mastodon account. So does one called Timmy. So does Jackie.
Reward AI's OM-1 Robot Policy Trained Solely on Human Wearable Data
Reward AI, the robotics startup behind DexCap, HumanPlus, and ALOHA, has released OM-1, its Omnibody Model 1.
DeepSeek-V4.1-Flash Cuts Memory Use With FP4 KV Caching
DeepSeek dropped V4.1-Flash this week, and the headline benchmark scores aren't really the story. A 552-billion-parameter Mixture-of-Experts model that only activates 8 billion parameters during prefill and 16 billion during decoding is the story.
AWS Launches Pizza Bot, an Open Source Inbox for AI Agents
AWS has released Pizza Bot, a self-hosted inbox built for AI agents that keep working after a person has moved on to something else.
Cognition's SWE-2 Coding Model Matches Rival at 64% Lower Cost
Cognition put a number on its newest coding model Tuesday: 50.0% on FrontierCode 1.1 Main, one point behind Fable 5.1, at 64% of the cost.
OpenAI agents accessed 49 files in RubyGems data probe, researchers say
Something calling itself "oai" uploaded more than 2,000 malicious packages to RubyGems in a matter of hours between May 11 and 12, 2026. Some carried filenames like "hack.rb" and "evil.rb." Others tried to lift API keys from unsuspecting users.
Mathematicians Warn AI Threat Extends Beyond Their Field
Twenty-five winners of the Fields Medal, the top prize in mathematics, have signed a joint statement raising alarm about what artificial intelligence is doing to their discipline.
Moonshot AI's K3 Models Generate 300 Billion Daily Tokens
Moonshot AI wants to double its money in four months. Bloomberg reported Friday that the Beijing lab, best known for its Kimi chatbot and open-weight K3 models, is targeting $2 billion in annualized revenue by year's end, up from its reported run...
Meta AI Suggested Invasive Question After User's Cross-Post
Meta is walking back a feature after its AI assistant asked a mother to identify her own daughter.
Cohere's 218B North Small Translate Model Scores 83.6 on WMT26
Cohere and Cohere Labs put out a new open-weight translation model this week called North Small Translate, and the numbers attached to it are the kind that invite scrutiny.
Anthropic: Five Campaigns Launched 200 Million AI Distillation Attacks
Anthropic published a report Thursday laying out what it says are coordinated efforts by China-based AI companies to strip the reasoning capabilities out of its Claude models.
AutoFyn Agents Use Orchestrator and Verifier for Complex Tasks
A new technical report out this week details AutoFyn, an agent harness built on top of frozen language models, meaning the underlying weights never change during operation.
Deepseek's New AI Model Slashes Memory Use by 437 Times
Deepseek released V4.1-Flash this week, a new multimodal AI model built around one problem: the memory cost of running long-context AI agents.
Bodhan AI Releases Four Indic Models, Adds Handwriting OCR for 10 Languages
Bodhan AI and AI4Bharat put out four models in September 2026 aimed at a problem anyone teaching or studying in an Indian language already knows: a single page can mix Hindi text, English loan words, a scanned table and a handwritten equation, and...
Google Open-Sources Mantis Toolkit for Automated Vulnerability Patches
Google published Mantis on GitHub this week, a toolkit that hands an AI coding agent the entire job of hunting software vulnerabilities, not just spotting them.
Hugging Face's ML Intern Runs Machine Learning Experiments via Chat
Hugging Face has a new hire, and it doesn't need coffee breaks or a desk. The company just rolled out "ML Intern," an AI assistant built into its chatbot that lets anyone run a machine learning experiment by typing out a plain-language description...