Baseten's Base Labs teams up with Hugging Face and Goodfire on open-weight AI safety standard

Baseten announced on Wednesday a new safety infrastructure initiative led by Base Labs, the research group it launched earlier this year, in partnership with model-hosting platform Hugging Face and interpretability specialist Goodfire AI.

The partnership aims to develop tools and methods for evaluating and continuously monitoring the safety of open-weight AI models. Base Labs intends to publish these methods and position them as a standard where safety is built directly into how models are trained and deployed, rather than added afterward.

The abliteration problem

The announcement arrives amid ongoing industry debate over the risks posed by open-weight models. These models can be made dangerous through a technique called abliteration, which strips out built-in safety guardrails. According to Baseten, Hugging Face currently hosts more than 6,000 models that have undergone this process.

In a post on X, Baseten argued that openness is an advantage for AI safety, since it offers greater visibility into model behavior and more ways to translate safety research into concrete, transparent controls, compared with closed-source approaches.

The companies have not disclosed technical details of how the partnership will operate. Goodfire, responding to Baseten's announcement, said safety must be built into open models and provided by those who serve them — suggesting Goodfire, which specializes in explaining how AI models make decisions, would take the lead on that front.

Company backgrounds

Baseten, an AI inference provider, raised a $1.5 billion Series F round in June, pushing its valuation to $13 billion. Goodfire AI raised a $150 million Series B round earlier this year, led by B Capital, to advance its model interpretability platform.

Baseten is calling on the broader developer community to contribute to the emerging safety framework.