News

Thomas Wolf announces Hugging Face Open Alignment team

In a 10 September 2026 post, Thomas WolfDealroom has a profile for this one. Try Dealroom →, co-founder and Chief Science Officer of Hugging Face, announced that the company is starting an Open Alignment team. The team will work on safety and alignment for open models, including cybersecurity, extending Hugging Face’s public focus on open-source and open-weights artificial intelligence.

What Wolf is calling for: The announcement accompanied Wolf’s disclosure that he had published an opinion article in the Financial Times about the OpenAI–Hugging Face incident and its follow-ups. He argued that the field needs substantially more transparency and research on the safety of open models. The post presents the new team as a practical contribution to that work, rather than treating alignment as a question limited to closed frontier-model developers.

Why it matters: Open models can be downloaded, adapted and deployed by a broad community, making safety research, evaluation and cybersecurity relevant beyond the organisations that train the largest systems. The announcement places Hugging Face’s open-model ecosystem in the wider debate about how to improve the transparency and accountability of increasingly capable AI systems, while leaving the team’s membership, research agenda and initial outputs to be detailed in future updates.

Read more: Thomas Wolf on X

More top stories