Exaggerating or inventing AI capabilities to market basic heuristics, commodity API wrappers, or concealed manual human labor as proprietary artificial intelligence.
AI washing is the deceptive practice of marketing conventional software, basic heuristics, thin third-party API wrappers, or concealed manual labor as advanced, proprietary artificial intelligence. Borrowed from “greenwashing,” the phrase names the widespread corporate reflex to slap an AI label on standard products to capture hype, inflate enterprise valuations, and close sales.
The tell is the gap between the promotional narrative and the actual system architecture. In 2019, an MMC Ventures study across European startups revealed that 40 percent of companies categorized as AI startups showed no material evidence of machine learning in their core product. By the mid-2020s, regulatory enforcement followed: in March 2024, the SEC issued $225,000 and $175,000 penalties against investment advisers Delphia and Global Predictions for falsely claiming their forecast models used AI. Regulators broadened scrutiny with the FTC’s “Keep your AI claims in check” guidance and Operation AI Comply sweep, while orders against drive-thru automation companies like Presto Automation exposed systems marketed as autonomous AI that relied on secret off-site humans intervening in most transactions.
Transparent human oversight is not AI washing. Having engineers review outputs, moderate safety edge cases, or verify high-stakes answers is standard human-in-the-loop engineering. AI washing occurs specifically when human involvement or off-the-shelf plumbing is intentionally hidden to manufacture the illusion of autonomous proprietary intelligence.