THE CRUNCH
Base Labs, the research group spun out of Baseten, has announced a partnership with Hugging Face and Goodfire to develop and publish methods for training and monitoring open models. The collaboration aims to advance open-weight AI safety practices by sharing technical approaches publicly. The initiative follows a recent call from Y Combinator’s Garry Tan for frontier models to be treated as a public good, suggesting
Goodfire, a startup specialising in interpretability, will contribute its tools for analysing model behaviour, while Hugging Face will host the resulting resources. The partnership is notable for its focus on making safety techniques accessible rather than just releasing model weights. This approach could help the broader AI community build and audit open models more effectively.