Hugging Face says its Open Alignment Initiative, led by co-founder Thomas Wolf, seeks “to be part of the ‘embedded evaluators’ program that Amodei” committed to
It's now clear that alignment is critical and won't be solved behind the closed doors of a handful of frontier labs. So today we're launching the Open Alignment Initiative, led by @Thom_Wolf @huggingface and asking to be part of the “embedded evaluators
Context & Ripple Effects
OpenAI had already made alignment a formal product-and-governance function through deliberative alignment for its reasoning models and a Collective Alignment team focused on public input. In 2026, OpenAI and Anthropic also backed the “Pacing the Frontier” initiative, placing safety commitments closer to an institutional agenda for frontier-model development.
Hugging Face is pressing for that agenda to include an open-model platform as an evaluator rather than leaving assessment to a small group of frontier labs. Public reaction focused on whether credible auditing requires participants outside firms’ existing networks.
First-order effects
- Hugging Face and Thomas Wolf gain a formal vehicle to seek a role in Amodei’s embedded-evaluators commitment, putting the initiative’s open-alignment work before a frontier-lab safety program.
- Amodei’s evaluator commitment acquires a concrete external applicant, making participation criteria consequential for Hugging Face and other prospective independent evaluators.
Second-order effects
- Admission of Hugging Face would turn embedded evaluation from a lab-centered safeguard into a channel for outside alignment organizations, raising the value of transparent evaluator-selection standards.
- OpenAI’s earlier public-input and deliberative-alignment efforts provide a competing reference point for how frontier labs can demonstrate that safety work extends beyond internal model teams.
Third-order effects
- If external groups receive meaningful evaluator access, alignment oversight may develop into a distinct institutional layer between model developers and the public rather than remaining solely an in-house research function.
- The contest will shift from whether labs make safety commitments to who can inspect, test, and credibly represent broader interests within those commitments.
The trend: Frontier AI safety is moving toward institutionalized oversight in which access for external evaluators becomes a measure of the credibility of lab-led alignment commitments.