Open weights are not open source: why AI's favorite label is under dispute
StanfordHAI · x · 2026-09-17
The Register examines the industry's loose use of "open": publishing model files to Hugging Face is open weights, not open source. Open weights let you self-host, fine-tune, and avoid vendor lock-in, but genuine open source requires the freedom to inspect, reproduce, modify, and redistribute. Stanford HAI's Landay: "Open weights answer 'Can I run this?' Open source answers 'Can I trust this, improve it, and build the next thing on top of it?' Right now almost everyone is answering the first question but nowhere close to the second."
Related event: Open weights ≠ open source: AI industry's favorite label under fire(2 posts)→
More from Companies & People
- Anthropic and OpenAI pledge to embed third-party safety evaluators, but true independence remains an open question — RebeccaBellan · 2026-09-17
- Baseten partners with Goodfire and ba labs on safety and interpretability for open models — JJitsev · 2026-09-17
- Educator: use AI to make students stop and think, not grab the fastest answer — azed_ai · 2026-09-17
- Benioff at Dreamforce: Enterprise AI breakthrough comes from the platform, not the model — sanderssays · 2026-09-17
- Buyer trust for AI services dissected: 'enterprises buy from vendors they trust' is a useless cliche — jonerp · 2026-09-17
- Thom Wolf partners with Baselabs, Baseten and Goodfire on open-model safety research — Thom_Wolf · 2026-09-17