NYC's AI auditing law offers lessons but lacks teeth
-
New York City's law requiring algorithmic fairness audits offers lessons for federal agencies implementing AI. However, compliance appears shockingly low so far.
-
The law's vague definitions and criteria allow employers to creatively lawyer around it. Future regulations need broader definitions based on systems' purpose, not just technical specs.
-
The law does not address the role of platforms and vendors, leaving accountability unclear. Regulations should align incentives and obligations across the AI ecosystem.
-
Without guidance on permissible bias levels, audits create legal risks. Safe harbors are needed so good faith auditing occurs industry-wide.
-
Requiring public audits is toothless without centralized repositories. Accessibility and enforcement depend on easy availability of audits.