Generative AI systems often produce false and defamatory statements about real individuals, but
traditional defamation law falters when the speaker lacks a mind. New York Times Co. v. Sullivan
requires proving actual malice – knowledge of falsity or reckless disregard for truth, but
algorithms have no ability to know about falsity or act recklessly. To address the gap, this
research proposes a doctrinal framework – redistributed actual malice – that maintains Sullivan's
constitutional protections by shifting fault inquiries from algorithmic outputs to human actors
related to AI outputs. Drawing on Eugene V olokh's notice-based liability, Nina Brown's
product-liability concept, and traditional defamation doctrine, the framework allocates
responsibilities this way: developers face heightened liability after receiving notice of specific
false outputs; deployers face context-dependent negligence standards; end-users face traditional
actual malice adapted to AI's documented hallucination risks. This three-tier structure maintains
the public figure and private figure distinction and demonstrates that actual malice's
constitutional function survives technological disruption, through systemic reallocation, instead
of doctrinal abandonment.
Publication Date: 2026-06-20