AI Washing

AI washing is overstating or fabricating the role artificial intelligence plays in a product, service or delivery model, usually to win buyers, investors or headlines. The name borrows from greenwashing, and the sales pattern behind it runs much the same way.

The claim outruns the technology. A vendor says “AI-powered” when the system is a rules-based script — a lookup table, or a person typing behind a chat window.

Sometimes the exaggeration is marketing sloppiness. Sometimes it’s fraud, and regulators now treat it that way.

For outsourcing buyers, the stakes are practical. You’re pricing a contract against productivity gains that may not exist, and the gap only shows up months into delivery.

Key takeaways

• AI washing means claiming more artificial intelligence capability than a product actually has.
• The United States Securities and Exchange Commission (SEC) charged two advisers over false AI claims in March 2024.
• The Federal Trade Commission (FTC) followed with five enforcement actions in September 2024.
• In outsourcing, the tell is usually a pilot that never reached production.
• Ask which decisions the model makes unaided — that question breaks most inflated claims.

How it works

AI washing works because buyers can’t see inside the product. The label sits on the outside, the mechanism stays hidden, and few contracts define what “AI-powered” actually means. That gap between claim and code is where the practice lives.

Vendors rarely lie outright. They stretch. A single generative AI feature bolted onto legacy software becomes “an AI-native platform” in the deck.

A proof-of-concept that ran for six weeks becomes “deployed at scale”. A human quality-review step disappears from the diagram entirely.

The most common versions follow a small number of shapes:

Detection is a question of specificity. Vague claims survive vague questions and collapse under precise ones.

Ask which decisions the model makes without a human. Ask for the human-in-the-loop ratio. Ask what the system does when it produces a bad output, and ask to see the AI audit trail that proves it.

A structured AI vendor evaluation turns those questions into scoring criteria rather than hopeful conversation.

Examples

Regulators moved on AI washing across 2024, and the published cases show exactly what crosses the line from optimistic marketing into enforceable deception. Two actions set the reference points buyers and providers now work from.

SEC, March 2024. The SEC’s 18 March 2024 enforcement release settled charges against investment advisers Delphia (USA) Inc. and Global Predictions Inc. over false and misleading statements about their purported use of AI.

Toronto-based Delphia claimed from 2019 to 2023, in SEC filings — a press release and on its website, that it used AI and machine learning incorporating client data. It did not have those capabilities.

San Francisco-based Global Predictions falsely claimed in 2023 to be the “first regulated AI financial advisor”.

The firms paid $400,000 in total civil penalties — $225,000 and $175,000 respectively. Both were censured and ordered to cease and desist, charged under the Marketing Rule barring any untrue statement of material fact in an advertisement.

Then-SEC Chair Gary Gensler put it plainly: “Investment advisers should not mislead the public by saying they are using an AI model when they are not. Such AI washing hurts investors.”

FTC, September 2024. The FTC’s Operation AI Comply announcement of 25 September 2024 brought five law-enforcement actions against operations using AI hype or selling AI tools usable in deceptive ways.

Cases covered a company promoting an AI tool for creating fake reviews, a company selling “AI Lawyer” services, and companies claiming AI could make consumers money through online storefronts.

DoNotPay claimed to offer “the world’s first robot lawyer”. Per the complaint, it ran no testing on whether its chatbot output matched a human lawyer’s work and hired no attorneys.

It agreed to a proposed order paying $193,000 and notifying subscribers from 2021 to 2023 about the service’s limitations.

The outsourcing version. A provider pitches AI-augmented BPO delivery, then staffs the account conventionally and calls the macros intelligent. The billing reflects automation the client never receives.

Related terms

AI washing sits inside a cluster of terms covering vendor honesty, model behaviour and buyer diligence. These related entries explain the controls and failure modes that surface most often when an inflated AI claim finally meets production reality.

• Greenwashing: the environmental-claims practice that gave AI washing its name and its playbook.
• AI Vendor Evaluation: a structured scoring process for testing supplier AI claims before signing.
• AI Hallucination: confident but false model output, often hidden by vendors selling accuracy.
• AI Pilot To Production: the transition stage where overstated deployments are usually exposed.
• AI Guardrails: the controls limiting what a model can do and say in live operations.

FAQ

Is AI washing illegal?

It can be. Both the SEC and FTC treated false AI claims as actionable deception in 2024, using existing advertising and consumer-protection rules rather than new AI-specific law.

How is AI washing different from normal marketing hype?

Hype exaggerates benefits; AI washing misstates a fact about what the technology is. Saying a model is “powerful” is opinion, while saying it exists when it doesn’t is a false statement.

What questions expose AI washing in a BPO pitch?

Ask which decisions the model makes unaided, what the human-in-the-loop ratio is, and what happens when the system produces a bad output. Then ask to see the audit trail behind each answer.

Can a provider be AI washing without meaning to?

Yes, and it happens often when sales teams repeat internal shorthand about a pilot that engineering never moved into production.

Browse vetted providers in the Outsource Accelerator directory to compare AI claims against documented delivery.