Hugging Face Launches Open Alignment Push

Episode 1 · Hugging Face Co-founder Thom Wolf: Alignment Can't Be Solved Behind Closed Doors (2026-09-10, 2 posts)

Hugging Face co-founder Thom Wolf argues that alignment is a critical unsolved problem for AI's future and cannot be solved behind closed doors at frontier labs, calling for open safety and alignment research and open-source models for defensive purposes.

• Hugging Face's Thom Wolf calls for open safety research and open-weight AI models for defense — Thom_Wolf · 2026-09-10
• Wolf: Alignment Won't Be Solved Behind Closed Doors at Frontier Labs — Thom_Wolf · 2026-09-11

Episode 2 · Hugging Face Launches Open Alignment Team, Seeks to Join Anthropic's Evaluation Process (2026-09-11, 6 posts)

On September 11, Hugging Face co-founder Thom Wolf announced two moves: an op-ed in the Financial Times on the OpenAI/HF incident and its aftermath, and the creation of an internal Open Alignment team at Hugging Face dedicated to safety and alignment research for open models, including cybersecurity, with calls for greater transparency and more research investment. CEO Clement Delangue added that the initiative will apply to join Anthropic's third-party evaluation process. Multiple bloggers (reachvb, MaziyarPanahi, CFGeek) relayed the news consistently.

Confirmed

• Thom Wolf published an FT op-ed on the OpenAI/HF incident
• Hugging Face formed the Open Alignment team, led by Thom Wolf, covering safety and alignment of open models including cybersecurity
• CEO Clement Delangue announced the initiative will apply to join Anthropic's evaluation mechanism
• The team argues alignment cannot be solved by a few frontier labs behind closed doors, calling for transparency and more research investment

Why it matters

Anthropic CEO Dario Amodei had argued the industry should slow frontier AI development and pledged permanent, staff-level access for third-party evaluators to verify safety measures. Delangue's response and the Open Alignment Initiative represent a substantive move by the open-source camp on safety: bringing alignment research into the open ecosystem while plugging into the safety-verification pathway led by closed labs, creating a bridge between the two approaches.

• Hugging Face Launches Open Alignment Team, Wolf Demands 100x More Transparency — Thom_Wolf · 2026-09-11
• Hugging Face launches Open Alignment team for open model safety and alignment — reach_vb · 2026-09-11
• Hugging Face launches Open Alignment team for open-model safety after FT op-ed — CFGeek · 2026-09-11
• Thom Wolf publishes FT op-ed, launches Open Alignment team at Hugging Face — MaziyarPanahi · 2026-09-11
• Hugging Face launches Open Alignment Initiative, responds to Dario Amodei's frontier pacing plan — huggingface · 2026-09-12
• Hugging Face Launches Open Alignment Initiative, Applies to Anthropic's Evaluator Program — TheMoonMidas · 2026-09-12