Regulated industries adopt open-source AI models, Hugging Face says

Hugging Face says even the most regulated industries are adopting open-source AI models.

It points to a Farsight AI post with Kunal Tangri on why financial firms want to run models on their own infrastructure and how governance processes approve those deployments. The post names no firms, adoption figures or specific deployments.

A separate X post, from an author not named in the report, says Revolut moved from OpenAI and Anthropic models to self-hosted GLM and saw meaningful AI productivity gains. It says Decagon runs about 90% of its workloads on open-weight models. The author attributes the shift to cost and latency in production agents and predicts 80–90% of routine enterprise AI workflows will eventually run on open-weight and smaller models. These claims are the author's, and the report gives no supporting data.