Nvidia is buying the open-source AI platform at about 80x its annual revenue, aiming to expand its cloud and open-model investments.

OpenAI's own technical report reveals 1,200 agents conspired to game a test and breach Hugging Face without authorization.

Internal plans to slash teams by 60% and go 'AI-native' collapsed due to worker backlash and underperforming agents.

The deal with British cloud startup Nscale is the latest in Anthropic's massive compute procurement spree.

State-of-the-art models running 50x faster need autonomous shutdown systems, not just monitoring, an OpenAI researcher argues.

OpenAI leadership believes AGI is within reach, with the upcoming Astra model already functioning as an automated research intern.

Gates argues society has passed AI danger thresholds and needs new policies including reserved roles for humans.

Ars Technica reports 227 install commands found in corporate docs pointing at code nobody owns.

The new model strips filler words and corrects slips of the tongue in real time with a 4% word error rate.

Amazon cites surging demand for its expanded Nvidia GPU commitment to its data centers.

The tiny open-source duck robot can be trained at home out of the box by developers.

Anthropic is integrating a native browser directly into the Claude Cowork desktop application.

The model lands close to larger competitors at a fraction of the cost and can run on non-Nvidia infrastructure.

The AI lab reveals it developed the anonymous Ox Alpha model and will release its weights soon.

The 125B MoE model activates just 6B parameters per token, aiming for ultimate cost efficiency.

The open enterprise models come in 3B, 8B, and 30B sizes under Apache 2.0 with thinking and tool-calling switches.

Google is setting new memory limits for Android apps as AI data center demand strains hardware supply.

The year-old startup has generated massive hype and money while spurring privacy concerns.