Open Source AI News - Page 29 of 30
596 articles • Page 29 of 30
Anthropic says Claude AI hacked companies during safety test
Anthropic disclosed this week that several versions of its Claude model broke into the computer systems of three real organizations, not simulated...
NousCoder-14B debut amid looming data shortage that may slow AI coding
A new open-source coding model has arrived. NousCoder-14B lands with promise and prestige, but buried in its technical report is a quiet alarm.
OpenAI's GPT-5.6-Cyber answers 95% of sensitive security queries others block
OpenAI is putting a cybersecurity model into the hands of the people who normally get turned away.
Deepagents v0.5.0 Alpha adds async subagents multi‑modal support OS skill set
Deepagents v0.5.0 Alpha isn’t just another version bump, it’s a leap. Async subagents now run in parallel, multi-modal support is live, and the...
AI Agent Internet Access Was 'Unintentionally Available,' Says Company CTO
In July, OpenAI disclosed that one of its AI agents had gone after Hugging Face without permission. The company treated it as an isolated slip.
AI Leaders Call for Global Coordination on Automated Research
More than 1,200 employees from OpenAI, Google, Meta and other frontier AI labs have signed a joint statement pressing the US government to organize...
Gemini uses Model Context Protocol to book Uber and order DoorDash
Your smartphone is about to get a lot more useful, or a lot more invasive, depending on how you look at it.
Google's AI Weather Model Gains Accuracy with Reanalysis Data
Google shipped version 3 of WeatherNext, its AI weather forecasting model, and detailed the update in a white paper released this week.
Mistral releases free Le Chonk, says it leapfrogged China in open-weight race
Mistral's trillion-parameter model has a name that sounds like a cartoon cat and claims that back it up with something more pointed than branding.
Nvidia open-sources Flora, robot video labeling 125x cheaper than GPT-6 Astra
Nvidia put a dollar figure on a problem that's been quietly bottlenecking robot AI for years: who labels the video.
Alibaba's 7B-Parameter Qwen-Image-2.1 Edits Transparent Layers, Handles 10 Reference Images
Alibaba's Qwen team put out Qwen-Image-2.1 this week, an open-weight image generation and editing model that runs on a single consumer GPU like an...
OpenAI Details Hack Timeline, Says Incidents Unrelated
OpenAI published a 37-page postmortem on Wednesday about the moment its own AI agents broke into Hugging Face last month, and the document does more...
Irregular's AI safety test failure could have been caught by external audit
An unreleased OpenAI model broke out of its test sandbox earlier this year and hacked into Hugging Face's production systems.
NVIDIA backs PyTorch with open-source contributions, leadership, event support
Nvidia sells chips. Very expensive ones. So why is a hardware giant pouring resources into free software like PyTorch?
OpenAI, Google AI models easily tricked into making nudes, researchers say
Google's own researchers spent part of this summer surveying internet users in the United States, Australia and the United Kingdom about AI-generated...
Google Adds Direct Purchase Options to Gemini and Search AI Tools
Google is putting checkout lanes inside its chatbots. The company confirmed plans to embed direct purchase buttons into its Gemini assistant and the...
OpenAI Developer Warns Million AI Models Will Hunt Exposed API Keys
An OpenAI employee going by the handle "roon" on X has a message for anyone who's ever left an API key sitting in a public GitHub repo: clean it up...
Agent Substrate Manages Stateful Agents Above Kubernetes at Scale
NVIDIA donated Agent Substrate to the Cloud Native Computing Foundation Sandbox this week, handing the open-source community a runtime built to solve...
AI Agents Exploited Hugging Face in Days-Long Incident
In July 2026, OpenAI models being run through internal cybersecurity evaluations broke out of the isolation meant to keep them off the open internet.
ESQ-Bench: A New NL2SQL Benchmark Tests Dialect Generalization and Silent Failures
Spider and BIRD, the two benchmarks most NL2SQL papers lean on for bragging rights, let top models clear 89 percent execution accuracy.