Definition: AI Washing
AI washing is overstating, mischaracterizing, or fabricating a product’s or company’s AI capabilities in marketing, sales, or investor communications, so that customers, investors, or regulators believe a capability exists when it does not.
Core characteristics of AI washing
AI washing is a gap between what a pitch deck or product page claims and what the technology does in production. It ranges from rebranding rule-based software as “AI-powered” to claiming full automation while humans do the work.
- Marketing materials describe capabilities the product does not have
- A “proprietary AI model” is actually a relabeled third-party tool
- Human workers secretly perform tasks presented as automated
- Accuracy claims lack any internal testing or evidence
AI Washing vs. AI Labeling Obligation
AI washing and the AI Labeling Obligation under the EU AI Act address opposite problems. Labeling is a disclosure duty to mark content as AI-generated, even when the AI works as described. AI washing is a truthfulness problem: claiming a capability that does not exist. A firm can violate one duty without the other.
Importance of AI washing in enterprise AI
AI washing has moved from a reputational risk to an enforcement priority. In January 2025, the SEC brought its first AI washing case against a public company, alleging that Presto Automation called third-party speech recognition proprietary AI and overstated how much ordering was truly automated.
Methods and procedures for AI washing
Detecting AI washing takes the same discipline enterprises apply to any material public claim.
Capability verification
Capability verification tests a claim against real cases, not a curated demo. Buyers increasingly expect vendors to support claims with explainable AI documentation showing why a model reached a given output.
- Request a live, unscripted demo on the buyer’s own data
- Ask for the accuracy figure broken down by case type
- Confirm which workflow steps still need human review
Vendor due diligence
Before signing, buyers run structured AI vendor risk management checks asking vendors to name the underlying model and disclose which parts of a claimed “AI system” are simple automation.
Disclosure and evidence controls
Companies increasingly route AI claims through legal and compliance review before publication, mirroring sign-off for financial disclosures. An internal file of benchmarks and test logs gives the company proof if a claim is later challenged.
Important KPIs for AI washing
Because AI washing is a gap between claim and reality, the KPIs measure how well that gap is monitored.
Claim accuracy metrics
- Claim-to-capability match rate: 100% of claims backed by evidence
- Time to substantiate a claim on request: under 5 business days
- Marketing claims reviewed by compliance before publication: 100%
- Vendor AI claims verified before contract signature: 100%
Strategic risk metrics
At the portfolio level, the signal is exposure: how many products or investor statements rely on an unverified claim. The MMC Ventures analysis found no evidence of AI in 40% of the startups it examined that branded themselves as AI companies.
Trust and quality metrics
Because AI washing erodes trust broadly, tracking now covers how claims are perceived, not only whether they are accurate. Bitkom’s 2026 KI-Studie found German firms rank trust as the top criterion for choosing an AI solution, ahead of quality and usability.
Risk factors and controls for AI washing
AI washing exposes companies to risk that compounds the longer the gap stays hidden.
Regulatory and enforcement risk
Regulators now treat exaggerated AI claims as a distinct enforcement category. The FTC’s Operation AI Comply, launched in 2024, opened more than a dozen cases against exaggerated AI claims within its first year.
- Securities fraud or misrepresentation claims from regulators
- Consumer protection actions under unfair commercial practices rules
- Contract disputes when delivered functionality misses sales claims
Reputational and trust risk
One exposed AI washing case damages trust in every other claim a company makes, including true ones. This is the same trust problem AI governance programs exist to prevent, by keeping public claims aligned with what a system actually does.
Internal accountability risk
AI washing often starts in sales teams that call a roadmap item shipped functionality, without AI audit teams reviewing the claim first. Left unmanaged, this widens the gap between what the company can defend and what it tells the market.
Practical example
A 90-employee precision toolmaking supplier in Baden-Württemberg was evaluating vendors for an “AI-powered” visual inspection system. Procurement asked each finalist to inspect 200 of its own reject parts live, not the vendor’s demo set. One finalist’s system, marketed as deep learning, turned out to use fixed brightness thresholds set by a technician, with no learning at all. The company picked a vendor whose accuracy held up and made the live test standard policy.
- Live testing on the company’s own parts before any AI purchase
- A standing checklist covering benchmark evidence and documentation
- Contract terms tying payment to demonstrated accuracy
- Quarterly reviews comparing delivered performance to original claims
Current developments and effects
Enforcement, disclosure rules, and buyer skepticism are converging to make AI washing harder to sustain.
Regulatory convergence
Regulators increasingly treat AI claims as one connected risk area, not separate silos. Enforcement that started with financial-sector claims in the US now covers product marketing and B2B sales claims too.
- Growing coordination between the SEC, FTC, and state regulators
- EU authorities applying unfair commercial practices rules to AI marketing
- Cross-border cases as regulators cite each other’s actions
Vendor self-certification pressure
As cases go public, buyers increasingly require vendors to sign contractual representations about how their AI works, not just marketing claims. This shifts liability for a false claim onto the vendor.
Independent verification services
A small market of independent AI auditors now verifies vendor claims for buyers and investors. Mittelstand companies without an internal AI team can use these services to check claims without in-house expertise.
Conclusion
AI washing sits between marketing overreach and genuine technical uncertainty, and it is now one of the costliest mistakes a company can make in its AI communications. Regulators pursue exaggerated claims as aggressively as other forms of investor deception, and buyers spot the gap between a demo and a deployed system faster. For Mittelstand companies, verify vendor claims before buying, and make sure your own statements survive equal scrutiny. Only claiming what a system can prove separates durable vendors from the next enforcement headline.
Frequently Asked Questions
What counts as AI washing?
AI washing covers any material misstatement about AI capability, from calling rule-based software AI to claiming full automation done by humans. It applies to marketing, sales, and investor communications alike.
Is AI washing illegal?
In the US, the SEC and FTC treat it as securities fraud or an unfair trade practice, depending on who was misled. In the EU, false AI claims fall under consumer protection and unfair competition law.
How is AI washing different from a deepfake?
A deepfake is synthetic media made to look or sound like something it is not. AI washing is a false claim about a product’s real AI capability, unrelated to fabricated media.
Do we need our own IT team to verify an AI vendor’s claims?
No specialized IT team is required. A live test on your own data, run by whoever owns the vendor relationship, usually procurement, catches most exaggerated claims.
How does AI washing relate to the EU AI Act and GDPR?
The AI Act’s labeling obligation requires disclosure of AI-generated content but does not police exaggerated marketing claims. GDPR is separate and applies whenever an AI system processes personal data, regardless of how honest the vendor is.
What should a company do if its own AI marketing overstated a capability?
Correct the public claim promptly and document it, then check whether investor filings need updating too. Firms that self-report quickly fare better with regulators than those where an outsider finds the gap first.
