Base Labs launches an open-weight AI safety partnership with Hugging Face and Goodfire
Base Labs, the research group Baseten spun up earlier this year, will develop and publish methods for training and monitoring open models.
Baseten, the research arm of AI inference provider Base Labs, announced on Wednesday the launch of a new open-weight AI safety partnership with Hugging Face and Goodfire AI. The joint effort aims to develop and publish methods for training and monitoring open models, addressing concerns about the safety of open-weight models. Hugging Face currently hosts over 6,000 abliterated models, which can become dangerous when safeguards are removed through a technique known as abliteration.
Base Labs, spun up earlier this year, will focus on creating transparent and integrated safety evaluation and monitoring infrastructure for open-weight models. The partnership's underlying goal is to build an open standard for open models, offering greater visibility into model behavior and enabling the creation of actionable safety controls.
Goodfire, specializing in model interpretability, is expected to play a key role in implementing the "built into" aspect of the partnership. Both Baseten and Goodfire have shown significant capital, with Baseten raising $1.5 billion in Series F funding in June, increasing its valuation to $13 billion, and Goodfire securing $150 million in Series B funding earlier this year.
The companies plan to open an invitation to the developer ecosystem to contribute to the framework, stating, "Together, we are building an ecosystem of open models that are safe and accessible to all."
Written by urgent.news from TechCrunch's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
