News reporting focused on AI and machine learning, covering the companies behind these technologies, their real-world applications, and the ethical concerns they raise. This includes areas like generative AI (large language models, text-to-image and video), speech tech, and predictive analytics.
The release of Kimi, a free open-source model from Chinese company Moonshot, has sparked public infighting among current and former Trump administration AI advisors. The model's capabilities rival those of paid US models from OpenAI and Anthropic, creating economic and political tensions. Disagreements have emerged over whether the government should restrict open-source competition or let market forces decide.
Current AI, a nonprofit founded in February 2025, is building open, public AI infrastructure to serve communities left behind by private AI systems. It has launched an offline device in India, allocated $3.2 million in grants, and released an open-source chatbot. The organization aims to create a public alternative to Big Tech's AI, modeled on the early World Wide Web.
Researchers at Google DeepMind have developed GenCeption, a model that repurposes a pre-trained video generator for classic computer vision tasks like depth estimation and segmentation. The system builds on an open-source video model from Alibaba and delivers results in a single forward pass, guided by text prompts. It trained on just a small set of synthetic videos and needs far less data than competing approaches. In benchmarks, GenCeption matches established specialized models and transfers its abilities to real-world footage and untrained categories like animals. The authors say this supports the contested idea that video generators can serve as the basis for universal world models in computer vision.
NVIDIA and Hugging Face have collaborated to bring production-grade distributed diffusion training to any Diffusers-format model on the Hugging Face Hub through the NVIDIA NeMo Automodel open-source library. The integration supports models like FLUX.1-dev, Wan 2.1, and HunyuanVideo, enabling full fine-tuning and LoRA adaptation without checkpoint conversion or model rewrites. The workflow uses YAML-driven configuration, pre-encoded datasets, and scalable parallelism from one GPU to hundreds.
General Compute, an AI inference cloud startup, secured a $400 million loan from Upper90, marking what may be the first deal to use inference-specific chips as collateral. The financing signals growing market demand for cost-efficient AI infrastructure that runs open-source models, as investors seek alternatives to expensive GPU-based systems.
Thinking Machines Lab, the startup founded by former OpenAI CTO Mira Murati, has released Inkling, an open-weights multimodal model with 975 billion parameters. The model leads U.S. open-source models on the Artificial Analysis Intelligence Index but trails top Chinese models. It shows strong agentic performance but has a 63 percent hallucination rate.
Simon Willison accidentally activated a pet in Codex Desktop and then created his own custom pet called Pedalican, a pelican on a bicycle. The process used GPT-5.6 Sol xhigh and gpt-image-2 to generate sprite assets. Willison documented the entire process, including generated images, sprite sheets, and GIFs, and shared the open-source skills on GitHub.
Cohere has launched Cohere Transcribe Arabic, an open-source 2-billion-parameter ASR model designed for Arabic speech recognition. The company claims it is the most accurate open-source Arabic speech-to-text system, outperforming Whisper Large V3 and the standard Cohere Transcribe model in benchmarks. The model handles dialect variety, code-switching, bilingual conversations, and specialized vocabulary.
FablePool is a new platform where strangers pool money behind a single instruction, and an AI agent builds the project in public, milestone by milestone. Funding targets are set by an AI planner, with contributions starting at $0.25. The site currently lists open projects including an open-source PID tuning library and a single-page website for Claude Shannon.
Anthropic has released an open-source reference implementation for autonomous vulnerability discovery and remediation using Claude. The framework includes skills for threat modeling, static scanning, triage, patching, and an autonomous pipeline designed for finding C/C++ memory vulnerabilities. It is based on learnings from partnerships with security teams and is intended as a starting point for building custom vulnerability finding pipelines.
Osaurus, an open-source Apple-only LLM server, lets Mac users switch between local and cloud AI models while keeping files and tools on their hardware. It evolved from a desktop AI companion called Dinoki and supports models like Llama, DeepSeek V4, and cloud providers such as OpenAI and Anthropic. The tool offers a secure sandbox, over 20 plugins, and has seen more than 112,000 downloads since launch.
Seattle startup CopilotKit has secured $27 million in Series A funding to advance tools for deploying AI agents directly in applications. The round, led by Glilot Capital, NFX, and SignalFire, supports development of its AG-UI protocol and enterprise toolkit. Backed by major providers like Google and Microsoft, the open-source protocol enables dynamic interfaces beyond text chats.