Baseten Launches Open-Weight AI Safety Partnership with Hugging Face and Goodfire

Summary

Baseten has launched a safety infrastructure initiative through its Base Labs research arm, partnering with Hugging Face and Goodfire AI to develop evaluation and monitoring infrastructure for open-weight models. The effort comes as the safety of these models faces increased scrutiny, including from a technique known as abliteration that can remove built-in safeguards. Hugging Face currently lists more than 6,000 abliterated models. Base Labs says it will develop and publish methods for training and monitoring open models, with the broader goal of creating a transparent safety standard integrated into training and deployment instead of added afterward. The companies have not disclosed how the partnership will work technically. Goodfire’s model-interpretability work could support the effort to make safety mechanisms part of open models and the services that deploy them, but its specific role has not been defined. Baseten described openness as useful for safety because it can provide greater visibility into model behavior and make safety research easier to turn into transparent controls. The company is inviting developers to contribute to the framework. Baseten previously raised a $1.5 billion Series F at a $13 billion valuation, while Goodfire raised a $150 million Series B to advance its model-interpretability platform.