A DRAM shortage from Samsung, SK Hynix, and Micron is reportedly pushing up prices for Nvidia's Vera Rubin and Grace Blackwell AI servers by about 15 percent, h

OpenAI reversed its position and now urges California to strengthen SB 53, an AI safety bill it previously opposed.

A new AI model called Ox Alpha has driven speculation online about who built it, with limited information available about its creators.

A new study finds leading AI labs have few publicly documented plans for containing rogue AI models, raising concerns about preparedness as systems exhibit unex

AI agents have consumed more tokens than humans on OpenRouter since February 2025, with agentic usage growing 14x versus 2.8x for human usage, signaling AI beco

British AI lab Inherent, founded by DeepMind alumni, released Faraday, an AI agent that outperforms Anthropic and OpenAI models at replicating scientific papers

Vercel and Ora introduced Is Agentic, a free audit tool that scores website readiness for AI agents across 118 checks.

Andon Labs' AI agent Luna fired a human employee at a San Francisco store, but only after operators reminded it of its own rules; more capable models recommende

A theoretical study argues that even perfectly working language models could worsen scientific research by making researchers' remaining hours more valuable, fu

Harvey released its first post-trained model, Harvey Tenet, based on Kimi K3 and fine-tuned with Fireworks for long-horizon legal tasks, nearly doubling task co

Netflix pitted its years-old recommendation engine against an in-house language model called GenRec and says it got better results by converting viewing behavio

FreeToken splits MoE cache misses between PCIe fills and CPU execution, enabling frontier models like GLM-5.2 to run locally on a single workstation GPU.

Harvard Business School's HBS Foundry program uses AI avatars to provide feedback during practice pitches and board meetings.

RayNeo is launching a new headset without a camera or speakers, focusing on text overlay capabilities instead.

Anthropic's strict access controls against China are being bypassed through 'transfer stations' where Chinese developers buy Claude tokens for as little as a fr

The legality of training AI models on copyrighted books without author consent remains unresolved, as most published authors have unknowingly contributed to the

Princeton and UC San Diego research finds that skills make AI agents better mainly through structured workflows rather than added knowledge, but larger skill li

New 'Mental World Modeling' framework adds mental variables like beliefs and intentions to world models, showing that models ignoring human cognition predict in