Baseten, Hugging Face Partner on Safety Standard for Open AI Models

Baseten has launched a new safety initiative through its Base Labs research arm, partnering with Hugging Face and Goodfire AI to establish safety evaluation and monitoring infrastructure specifically designed for open-weight AI models, according to TechCrunch.

The partnership, announced Wednesday, arrives as the AI industry grapples with a growing security challenge: abliteration, a technique that strips safety guardrails from open models. Hugging Face currently hosts more than 6,000 models that have been modified using this method, underscoring the scale of the vulnerability.

Building safety into model architecture

Base Labs will develop and publish methods for training and monitoring open models, positioning the work as a transparent standard built into how models are trained and deployed rather than added as an afterthought. The companies frame openness as a safety advantage, arguing it provides greater visibility into model behavior and more actionable controls than closed-source alternatives.

While technical implementation details remain undisclosed, Goodfire AI indicated the partnership's direction in response to Baseten's announcement: "Safety must be built into open models and provided by those who serve them." Goodfire specializes in model interpretability—opening AI's "black box" to explain decision-making processes—suggesting it will handle the architectural integration of safety features.

Why it matters

The abliteration problem exposes a fundamental tension in open AI development: the same accessibility that enables innovation also allows bad actors to weaponize models by removing safeguards. With thousands of compromised models already in circulation, reactive safety measures have proven insufficient. This partnership represents a shift toward proactive, architecture-level safety controls that can't be easily stripped away—a critical evolution as open-weight models gain enterprise adoption.

Well-capitalized safety push

Both lead partners bring substantial resources to the effort. Baseten, an AI inference provider, raised a $1.5 billion Series F in June at a $13 billion valuation. Goodfire AI secured $150 million in Series B funding led by B Capital earlier this year to advance its interpretability platform.

Baseten is issuing an open call for the broader developer community to contribute to the framework, aiming to build what it describes as "an ecosystem of open models that are safe and accessible to all."

The details were first reported by TechCrunch.