View: Open Models and Safety Can Coexist

ZhongRuiqi · x · 2026-09-01

The post discusses the feasibility of building AI models that are both open and safe. The author argues that open models are becoming increasingly important to avoid centralization and spread benefits, while safety remains critical. A linked post from the Thinkymachines team announces they are hiring safety researchers to work across the entire model development stack, including data filtering, capability evals, post-training safety, and red-teaming, with a focus on building evals for open-weight releases.

Original post →

More from Safety

Safety channel →