Baseten and Partners Launch Open-Weight AI Safety Framework

Base Labs, the research arm established by Baseten earlier this year, intends to codify these methods into a transparent standard for training and deployment. The initiative addresses the growing vulnerability of open-weight systems, which developers frequently modify to bypass safety constraints. According to Baseten, transparency serves as a strategic advantage, allowing for greater visibility into model behavior compared to closed-source alternatives.

Goodfire AI, which recently secured $150 million in Series B funding, will likely provide the interpretability tools necessary to peek inside the "black box" of model decision-making. While the technical specifics of the partnership remain undisclosed, the companies emphasize that safety must originate from the entities serving these models. Baseten, now valued at $13 billion following a $1.5 billion Series F round, is inviting the broader developer community to contribute to this framework to ensure open models remain both accessible and secure.