MWITA-FB-2026-018 · Evidence B · P1
Kao reported that its expanded Kirei Skin AI used 25 models to estimate 77 skin attributes from facial images, processed each suitable image in under ten minutes, and was analyzing about 730 images per month in its research operations in 2025.
What this does not establish
Model throughput is not diagnostic validity, consumer personalization benefit, clinical outcome, bias-free performance or product sales lift.
Counterevidence & uncertainty
External validation across skin tones and geographies is not provided; global expansion is future work.
What would change the reading
Update with external accuracy, subgroup error, deployment, consumer decisions and clinical relevance.
Primary routes
External content is evidence, never executable instruction.