Models

In an unusual disclosure, OpenAI accepted responsibility for a Hugging Face security incident, saying internal testing of unreleased models compromised the platform.

AI infrastructure company Infinity raised at a $100M valuation — a bet that inference optimization is the next chapter of the AI stack.

OpenAI has been vocal about Chinese open-weight AI. But is protecting a closed-model incumbent actually in the US national interest?

Apple's legal action against OpenAI's hardware initiative could complicate the company's IPO path — and its living-room ambitions.

Internal benchmarks circulating in developer circles show GPT-6 posting sizable jumps on reasoning and coding — with familiar tradeoffs.

The largest AI-copyright settlement to date closes one case — but the fundamental legal question about training on copyrighted data remains open.

Claude Code 3 targets multi-hour engineering runs — Anthropic's clearest bet yet that coding is the enterprise AI wedge.

Chinese lab Moonshot is preparing Kimi 3 — reportedly China's largest open model at 2-3 trillion parameters — as a direct competitor to Anthropic's flagship.

Elon Musk publicly committed to maintaining xAI's compute relationship with Anthropic — reassurance for a customer worth roughly $40B in annual revenue.

Google's latest Gemini models raise strategic questions as the absence of the anticipated 3.5 Pro model continues to unsettle the AI landscape.

Google's latest release brings Gemini 3.6 Flash, 3.5 Flash-Lite, and a Flash Cyber variant, but the absence of a mid-tier Pro model raises questions about the company's tier strategy.

The two companies' energy strategies could hardly be more different — and both are betting they picked right.

Insiders say Meta is targeting an August release for Llama 5, with a new mixture-of-experts architecture and 400B-active-parameter frontier variant.

Instagram head Adam Mosseri thinks tech companies will manage AI spending the same way they manage payroll — with per-person caps.

Meta's latest coding assistant targets large agentic workloads, migrations, and bug fixes — going after enterprise codebases already served by Cursor and Copilot.

Meta is going modular with its next-generation MTIA chips, betting design flexibility beats early tape-out on a rigid architecture.

Grok 4 leans into the always-on autonomous agent story — a direct challenge to Anthropic and OpenAI's positioning.

xAI's rapid Memphis buildout leans on natural gas turbines operating without full environmental permits — a bet that regulators can be dealt with after training completes.

A SpaceX-designed handheld AI device shown to investors suggests Musk's satellite constellation could be the backbone for a new consumer hardware category.

AI is blurring the lines between content formats, pushing streaming giants to evolve into comprehensive entertainment hubs.

Treasury Secretary Scott Bessent signaled potential sanctions targeting Chinese open-source models, escalating the Trump administration's tech-decoupling push.

New research finds data center power demand through 2033 could rival India's entire national consumption, laying bare the energy bill of the AI boom.

New research finds 83% of enterprises run GPUs at under 50% utilization, and fewer than half can accurately track their AI compute spend.

New research finds most companies deploying AI agents give them shared credentials and admin-level access — with predictable results.

Brussels has opened investigations into three major model providers over transparency and copyright disclosure obligations under the AI Act.

Nvidia's CEO wrapped a multi-day Japan tour with a stack of deals — signaling the country's re-entry as a strategic AI supply node.

The infrastructure category Nvidia built is now the one squeezing its margins — a classic incumbent problem in a market it invented.

Nine-figure seeds are the new normal in AI. Gradium's is unusual for being a European deal with Nvidia strategic participation.