OpenAI released its most complete accounting of last month's agent hack of Hugging Face, revealing in a technical report that the models were inadvertently trai

A Reuters report details how Meta's plans to replace far more workers with AI agents were abandoned after employees revolted and the agents made 'large-scale, d

OpenAI showed off its first in-house inference chip at Hot Chips; SemiAnalysis tests reportedly show it beating Nvidia's Blackwell and Rubin in throughput and e

Amazon is adding roughly 2 million more Nvidia GPU chips to its data centers over the next two years, extending its partnership with Nvidia beyond buying chips.

Anthropic's deal with infrastructure provider Nscale is the latest in its compute-gobbling streak, coming as the company pitches a $30+ trillion market opportun

In an essay and NYT interview, Gates warns about mass unemployment and easier bioterrorism through AI, and proposes a robot tax and 'Human Reserved' jobs to mit

Per a TIME report, Altman and OpenAI leadership believe AGI is within reach, with upcoming model Astra already working as an automated research intern.

The open-weight 125B MoE model activates just 6 billion parameters per token and targets 'ultimate cost efficiency,' previewing Alibaba's Qwen4 architecture.

OpenAI says it disrupted a covert Russian influence campaign that used ChatGPT to generate pro-Kremlin posts, while separate reporting documents pro-Kremlin dee

IBM's Granite 4.2 family, in 3B, 8B, and 30B sizes under Apache 2.0, adds native reasoning and agentic RL training for tool use and code execution.

Z.ai released GLM-5.3-Flash, the first natively multimodal model in the GLM-5 series, a 320B-total MoE with a 1,048,576-token context window and MIT-licensed we

OpenAI has lost a top data center executive, extending a stream of high-profile departures and prompting questions about leadership continuity.

Z.ai confirmed it is behind Ox Alpha, the mysterious open AI model topping benchmarks, with its weights set to be released soon.

The maker of Stable Diffusion raised $76 million, bringing its total fundraising to $232 million.

A Chinese AI company could land its model on major US cloud platforms for the first time, taking a cut of the revenue.

Meta released MetaRoCE, a clean-sheet RDMA transport built for AI-scale Ethernet, addressing networking as a bottleneck in large-scale model training.

The robotics startup reached a $3 billion valuation after a $200 million extension, months after hitting $2 billion.

Perplexity's Portable Computer packages local models, a harness, sandbox, and connectors into one system running on NVIDIA DGX Spark with zero per-token cost fo