Baseten launched a new safety infrastructure standard alongside its Base Labs research arm on Wednesday, partnering with Hugging Face and Goodfire AI to build evaluation and monitoring tools for open-weight models. The announcement arrives amid growing debate over the security of open-weight models, which can become dangerous if developers remove their safeguards through a rising technique known as abliteration. Hugging Face currently lists over 6,000 such models on its platform, highlighting the scale of the issue. Base Labs will develop and publish methods for training and monitoring these systems, framing its work as a standard that is transparent and built into the training process rather than added afterward. The company stated that openness provides greater visibility into model behaviour and more means of turning safety research into actionable controls than closed-source alternatives.
The partnership relies on significant capital behind both organisations. Baseten raised a $1.5 billion Series F in June to reach a $13 billion valuation, while Goodfire AI raised a $150 million Series B earlier this year to advance its model interpretability platform. Goodfire specialises in explaining how models make decisions, making it the likely candidate for the “built into” aspect of the safety framework. The companies have not disclosed technical details yet but are issuing an open call to the developer ecosystem to contribute to the framework.
- Baseten spun up Base Labs earlier this year as a separate research group.
- Goodfire’s platform focuses on opening AI’s black box to reveal decision logic.
- The initiative aims to make safety controls transparent and part of the deployment workflow.
