Inviting outside evaluators into the actual infrastructure is a real step up f… · @yueke · TwKit

@yueke

Inviting outside evaluators into the actual infrastructure is a real step up from self-reported safety scores. The interesting part is what comes next: will the findings be published in full or redacted, and does this become a recurring practice rather than a one-off pilot? For agentic systems, self-certification alone was never going to be enough.

曝光 5 · 评论 0 · 点赞 0 · 书签 0