The Securities and Exchange Commission has launched a systematic enforcement campaign against "AI washing" — the practice of companies and investment advisers making false or misleading claims about their use of artificial intelligence to attract investors, inflate valuations, or justify premium fees. The campaign has produced settlements, cease-and-desist orders, and ongoing litigation against firms that described "proprietary neural networks" that were spreadsheet formulas, "AI-driven portfolio optimization" that was manual stock-picking, and "machine learning platforms" that existed only in marketing materials. The enforcement effort reflects a regulatory pattern as old as the securities laws themselves: whenever a new technology generates speculative enthusiasm, the gap between what companies claim and what companies do becomes the SEC's primary investigative target.
The Three Forms of AI Washing
The SEC's enforcement actions have identified three distinct categories of AI misrepresentation, each with its own evidentiary signature and its own investor harm.
The first is capability fabrication — claiming to use AI technology that does not exist. In March 2024, the SEC settled charges against two registered investment advisers, Delphia Inc. and Global Predictions Inc., for materially misleading statements about their use of artificial intelligence. Delphia's marketing materials claimed its AI "put[] collective data to work to make our clients' money work harder for them," implying an operational machine learning system that analyzed aggregated client data to generate investment insights. Global Predictions marketed itself as the "first regulated AI financial advisor." Neither firm had deployed functional AI systems. Delphia paid $225,000 in penalties. Global Predictions paid $175,000. Both consented to cease-and-desist orders prohibiting future AI misrepresentations.
The penalties were modest. The precedent was not. The settlements established that the SEC treats AI misrepresentation with the same seriousness as any other materially misleading statement in SEC filings or client communications — and that the Commission has the technical capacity to distinguish between a genuine machine learning deployment and a marketing claim dressed in algorithmic language.
The second form is capability exaggeration — using AI in a limited, peripheral capacity while describing it as central to the firm's investment process or business model. A wealth management firm that uses a chatbot for appointment scheduling and describes itself as an "AI-powered wealth platform" is exaggerating the role of AI in its core function. The SEC has indicated through staff guidance that describing AI as integral to investment decision-making when it is used only for administrative or client-facing functions constitutes a misleading statement under the Investment Advisers Act of 1940.
The third form is buzzword insertion — the mechanical addition of "AI," "machine learning," or "deep learning" to 10-K filings, prospectuses, and investor presentations without any underlying technological implementation. The frequency of AI-related terms in public company SEC filings increased by approximately 250 percent between 2022 and 2025 — an increase that tracks the growth in AI-related search volume and investor interest rather than the growth in actual AI deployment. When a company's annual report states that it "leverages artificial intelligence across its operations" and the underlying reality is a pilot program that auto-categorizes expense reports, the gap between the filing and the fact is the gap the SEC is investigating.
How the SEC Investigates AI Claims
The investigative process combines traditional disclosure analysis with emerging technical competence. The Division of Corporation Finance reviews 10-K and 10-Q filings for AI-related claims and issues comment letters demanding specificity. A comment letter might ask: "You state that your platform 'utilizes proprietary AI algorithms.' Please describe the specific algorithms used, the data inputs processed, the decisions influenced by the algorithms, and the human oversight applied to algorithmic outputs." Companies that cannot provide substantive answers face referral to the Division of Enforcement.
For investment advisers, the SEC's Office of Compliance Inspections and Examinations conducts on-site inspections that now routinely include requests for AI documentation: source code repositories, model training records, validation and backtesting results, and evidence that the described systems exist in production and influence actual investment decisions. An adviser who claims to use a "proprietary neural network for portfolio optimization" must produce the network architecture, its training data provenance, its performance metrics, and audit logs demonstrating that the model's output was incorporated into portfolio construction decisions during the period covered by the marketing claim.
The evidentiary standard is not whether the AI claim is defensible in some narrow technical sense. It is whether a reasonable investor would be misled by the claim in the context in which it is presented. A company that uses natural language processing to summarize earnings transcripts for internal analysts but describes itself as "AI-native" in its investor presentation is making a claim that a reasonable investor would interpret as describing a fundamentally AI-driven business. If AI is peripheral to the company's revenue generation and competitive positioning, the characterization is misleading regardless of whether some AI technology exists somewhere in the organization.
How to Read Through AI Claims as an Investor
The investor evaluating a company's AI claims should apply three diagnostic tests that separate genuine deployment from marketing posture.
The first test is specificity. Does the company describe its AI capabilities in concrete, verifiable terms — identifying the type of model (transformer, convolutional neural network, gradient-boosted decision tree), the data inputs (structured financial data, unstructured text, satellite imagery), and the decisions the model influences (credit underwriting, inventory management, fraud detection)? Or does the company use vague formulations — "leveraging the power of AI," "AI-driven insights," "next-generation machine learning" — that describe aspiration rather than implementation? Specificity is the hallmark of genuine deployment. Vagueness is the hallmark of marketing.
The second test is investment consistency. Does the company's research and development spending support its AI claims? Training and deploying production machine learning systems requires significant compute expenditure, specialized engineering talent, and ongoing model maintenance. A company with $50 million in annual revenue that claims to operate cutting-edge AI but spends $500,000 on total R&D is not investing at a level consistent with genuine AI development. The gap between the claim and the capital allocation is itself informative.
The third test is organizational evidence. Does the company's workforce include machine learning engineers, data scientists, and MLOps specialists in numbers consistent with the scope of its AI claims? A single "Head of AI" and two junior data analysts do not constitute the engineering capacity required to build, train, validate, deploy, and maintain the enterprise-wide AI systems described in a presentation that implies algorithmic decision-making at every level of the organization. LinkedIn headcount data, job postings, and patent filings provide external verification of internal AI capacity that is independent of the company's own marketing narrative.
The SEC's AI washing enforcement campaign is not an indictment of artificial intelligence. It is the application of the most fundamental principle in securities law: investors are entitled to accurate information about what they are buying. The company that genuinely deploys AI to create competitive advantage should describe that deployment accurately and be rewarded by the market for it. The company that pastes "AI" onto a slide deck to capture the valuation premium that genuine AI companies command is defrauding its investors — and the SEC is coming for the slide deck.