Human-washing

Human-washing is marketing an AI system as having “human oversight” when the human’s actual role is decorative — a person positioned to watch or approve output they can’t meaningfully evaluate, providing the appearance of oversight with none of the substance. The tell is specificity: a designed system can say exactly what the human does, which cases reach them, and what triggers their yes; a human-washed one says “a human reviews everything,” which in practice means the human reviews nothing.

Why it matters: The washed version isn’t just weaker oversight — it’s a known failure mode with a forty-year paper trail: a person watching a mostly-right machine goes dull precisely as the volume climbs, then gets handed the hardest case cold. And the accountability never washes off — courts have already held companies responsible for what their AI tells customers, whatever the org chart says. The five words “there’s a human in the loop” are a claim to audit, not a guarantee to accept.