Base Labs Partners With Hugging Face and Goodfire to Build AI Safety Framework for Open Models

Base Labs, the research arm of AI infrastructure company Baseten, has launched a partnership with Hugging Face and Goodfire AI to develop safety evaluation and monitoring tools for open-weight AI models.

The initiative aims to create a more transparent safety framework for open models, with testing and monitoring methods built into the training and deployment process rather than added later.

Building safety tools for open AI models

The partnership comes as concerns grow around open-weight models, which can have their built-in safeguards removed through techniques such as ablation. Hugging Face currently hosts thousands of modified models with removed safety controls, highlighting the challenge of monitoring open AI systems.

Base Labs said openness can improve AI safety by helping researchers better understand model behaviour and develop clearer controls. The company shared its goals in an announcement on X.

Goodfire AI, which focuses on understanding how AI models make decisions, is expected to contribute expertise in model interpretability, while Hugging Face brings experience hosting and supporting open AI models.

Expanding AI safety infrastructure

Baseten has raised significant funding to expand its AI infrastructure business, while Goodfire has also secured investment for its model interpretability technology.

The companies are inviting developers and researchers to contribute to the framework as they work to create safer, more accessible open AI models.