Hugging Face's Irene Solaiman calls for standardized, open AI evaluations

Technology Oct 20, 2026

Irene Solaiman from Hugging Face isn't sold on the idea that built-in artificial intelligence (AI) evaluators can replace government rules.

Speaking at TechCrunch Disrupt, she pointed out that how we judge AI models isn't standardized and depends a lot on who's doing the checking.

She emphasized that for evaluations to mean anything, we need open access to models and a more reproducible and standardized system.

Solaiman questions 3rd-party reviewer neutrality

Solaiman also questioned if third-party reviewers are truly neutral, using Anthropic hiring Accenture as an example.

Without clear standards, it's tough to compare different AIs fairly or know which ones are actually safe.

She believes having one unified way to benchmark is essential if we want trustworthy and reliable AI development.