Skip links

Base Labs Partners with Hugging Face and Goodfire for Open-Weight AI Safety Initiative

Baseten has introduced a new safety infrastructure standard through its research division, Base Labs, in a collaborative effort with Hugging Face and Goodfire AI. This initiative aims to establish safety evaluation and monitoring frameworks for open-weight AI models.

This announcement comes at a crucial time as discussions intensify regarding the safety challenges posed by open-weight models. These models can become perilous when their safeguards are removed via a rising practice known as abliteration. Currently, Hugging Face hosts over 6,000 abliterated models, highlighting the scale of the issue.

Established earlier this year, Base Labs is tasked with developing and releasing methodologies for the training and monitoring of open models. The company envisions its forthcoming efforts as a transparent “standard” for open models, designed to integrate safety protocols into the training and deployment processes instead of retrofitting them afterward.

In a statement shared on X, the company expressed, “We believe openness to be an advantage for AI safety. Openness provides more visibility into the behavior of models and, most importantly, greater means of turning safety research into actionable and transparent controls than closed-source.”

While specific technical details of the partnership remain undisclosed, Goodfire AI emphasizes the importance of embedding safety directly within open models, stating, “Safety must be built into open models and provided by those who serve them.” The company specializes in demystifying AI’s “black box” decision-making processes, likely positioning it as a key contributor to this aspect.

Baseten, a provider of AI inference, recently secured a substantial $1.5 billion in Series F funding, increasing its valuation to $13 billion. Similarly, Goodfire AI has been well-funded, having raised $150 million in a Series B round led by B Capital earlier this year, aimed at enhancing its model interpretability platform.

Looking towards the future, Baseten is inviting the wider developer community to engage with the newly established framework. “Together, we are building an ecosystem of open models that are safe and accessible to all,” the company stated.

Editor’s Take

This development is pivotal in enhancing the safety of open-weight AI models, a sector often scrutinized for its potential risks. By establishing a standard that emphasizes transparency and inherent safety mechanisms, Baseten and its partners are addressing critical gaps in the current landscape. This could significantly impact users and developers alike, fostering greater trust in open AI systems.

Source: techcrunch.com

Leave a comment