How Base Labs, Hugging Face, and Goodfire are Making Models Safer.
“>Over 6,000 models on Hugging Face have had their safety filters lifted. Base Labs, Hugging Face, and Goodfire AI are partnering to build safeguard open-source models.
There are over 6,000 unfiltered (safety) models available on Hugging Face. They use a technique called “abliteration” to modify the weights of the model. Base Labs (the research arm of Baseten) has teamed up with Hugging Face and Goodfire AI. Together they aim to develop and share methods for training and monitoring open source models.
A good attempt at addressing a problem. If they have access to the weights and can modify them, then the vendor’s guarantee is nullified. This is just an announcement at this point with little detail. Let’s hope something materializes from it.
Source: techcrunch.com
Free course
Stop reading about AI — start building with it
The free Claude Code course: your first site, tool or game — no coding. No upsells, no cross-sells — nothing to buy here.
Start free →▌ Related guides
Author
Evgenii Arsentev
PhD · Chief Executive Officer, digital health
Articles · Latest articles