AI Washing
AI washing is the use of false, exaggerated, vague, or unsupported claims about an organization's or product's use, capability, autonomy, performance, or impact of artificial intelligence. A product can contain genuine AI and still be AI-washed if the marketing materially overstates what that AI does or what evidence supports the claim. The concept describes a mismatch between representation and substantiation, not merely the complete absence of AI.
Origin and context
A March 2017 InfoWorld article used AI-washing for marketing that made limited products sound more intelligent. A 2026 discourse analysis traced the same early use and documented the term's expansion across technology, finance, law, and general media. In 2024, the US Securities and Exchange Commission used AI washing in enforcement communications concerning investment advisers, while the Federal Trade Commission pursued unsupported claims about AI-powered professional services and commercial outcomes.
Why it matters
Inflated AI claims can distort purchasing and investment decisions, hide manual labor or conventional automation, and encourage reliance on systems that were not tested for the promised task. They can also expose organizations to securities, advertising, consumer-protection, contract, or sector-specific risk. A useful review connects each claim to a defined system, measurable capability, relevant test, operating conditions, and human contribution instead of treating the AI label as evidence.
Example
A vendor advertises an AI legal service as a substitute for a lawyer but has not tested equivalence and cannot substantiate the promised outcome. That is a stronger AI-washing signal than merely using an imprecise AI-powered label. A reviewer would request the model and workflow description, human-review boundaries, evaluation design, representative results, and limitations, then compare those materials with the exact claim and audience.
How it differs
Agent washing
Agent washing is a narrower subtype in which software is marketed as an autonomous or agentic system beyond its demonstrated behavior. It belongs as a reference-only companion under AI washing, not as an exact alias, because misleading AI claims also concern models, analytics, products, investment processes, and outcomes unrelated to agents.
Open-washing
Open washing misrepresents how open a model, dataset, or software project is. AI washing misrepresents AI use or capability. The practices can overlap when a provider exaggerates both openness and technical capability, but each has a distinct claim to test.
Maturity and evidence
Maturity is rated 4. The label has documented use since 2017, an established analogy, independent discourse evidence, and application in multiple US regulatory actions. It is not a single statutory offense with one global test. Whether a claim is unlawful depends on jurisdiction, materiality, audience, evidence, and the rules governing the speaker and transaction.
Limits and open questions
AI itself has contested boundaries, so the label can be used too broadly against ordinary simplification or good-faith product language. Technical novelty is not required for a product to provide value, and limited automation is not automatically deceptive. Reviewers should preserve the exact representation, identify the implied audience and decision, ask what evidence existed when the claim was made, and distinguish criticism from a legal conclusion. Current enforcement examples do not create a universal definition for every jurisdiction.
Related terms
References
- Artificially inflated: It's time to call BS on AIInfoWorld · 2017-03-03 · class B
- SEC Charges Two Investment Advisers with Making False and Misleading Statements About Their Use of Artificial IntelligenceUS Securities and Exchange Commission · 2024-03-18 · class A
- FTC Announces Crackdown on Deceptive AI Claims and SchemesUS Federal Trade Commission · 2024-09-25 · class A
- AI-Washing / KI-WashingDiskursmonitor · 2026-01-12 · class B
Last updated: 2026-09-04
This term is also covered in the Skills Atlas as ai ethics skill.
This term is also covered in the Skills Atlas as ai risk management skill.