Base Labs partners with Hugging Face and Goodfire to enhance AI safety for open-weight models
Baseten's research arm, Base Labs, has launched a new initiative in collaboration with Hugging Face and Goodfire AI aimed at improving the safety of open-weight artificial intelligence (AI) models. The partnership seeks to develop infrastructure for evaluating and monitoring these models as concerns grow over techniques like abliteration that can remove built-in safeguards. With more than 6,000 such modified models listed on Hugging Face's platform, there is a pressing need for robust safety measures. Baseten emphasizes the importance of integrating safety directly into model development rather than adding it later. The company aims to create transparent standards that can be adopted widely within the open AI community. Goodfire AI brings expertise in making AI decision-making processes more interpretable and will likely play a key role in embedding these safety features. Baseten recently secured significant funding, raising $1.5 billion for its Series F round, which has boosted its valuation to $13 billion. The company is now inviting developers worldwide to contribute to this open framework.
Topics: AI safety collaboration model monitoring open-weight models