AI Washing: From Marketing Hype to Material Liability
This analysis maps the full landscape – litigation, regulation, deal-making, and governance – and draws the implications for boards, executives, and institutional investors.
The Anatomy of a New Risk Category
For much of the past three years, “AI washing” was discussed primarily as a brand integrity problem – the corporate equivalent of greenwashing, concerning but manageable. That framing is now obsolete. As of early 2026, roughly 300 active AI-related cases are pending in U.S. courts, 17 class-action suits filed in 2025 alone account for nearly 10% of all new federal investor disputes, and the aggregate dollar value of investor losses associated with securities class action filings reached a record high last year. AI washing has crossed from marketing malpractice into the realm of material securities fraud.
The escalation is not incidental. It reflects a structural tension that has been building since 2022: the extraordinary valuations assigned to AI-adjacent businesses created enormous incentives for companies – large and small, tech and non-tech – to position themselves as AI-first, AI-powered, or AI-transformed. In many cases, those claims were not false in intent but reckless in execution, made without the governance infrastructure to substantiate them. Courts, regulators, and acquirers are now closing that gap at speed.
Two distinct forms of AI washing have crystallized in the legal record, and the distinction matters for how boards assess their exposure.
A Taxonomy of Deception
Product AI Washing – “Ghost Work” occurs when companies market products as driven by proprietary or autonomous AI when the core functions are actually performed by human labor – often offshore. The SEC’s January 2025 settlement with Presto Automation is the paradigm case: the company falsely claimed its AI processed restaurant orders when those tasks were handled by human workers. Innodata Inc. faces similar allegations – marketed as an AI-powered business while relying primarily on manual offshore operations. The fraudulent fundraising case against Nate, Inc. follows the same template: the founder allegedly raised over $42 million by claiming his app used AI to complete purchases, which were in reality executed by contract workers in the Philippines.
Operational AI Washing – is subtler and, arguably, more systemic. Here, management invokes AI as the public rationale for internal decisions – workforce reductions, restructurings, capital reallocations – that are in reality driven by financial distress, pandemic-era overhiring, or the need to free capital for genuine AI investment. A survey finding that 42% of HR managers admitted to citing AI implementation as the public justification for layoffs because “it sounded better than admitting to financial crisis” is a stark illustration of how normalized this practice has become. Deutsche Bank analysts explicitly flagged “AI redundancy washing” as a notable trend to watch in 2026. In the first quarter of this year, over 80,000 tech jobs were shed – the highest layoff count in three years – while AI was cited as a contributing factor in a substantial proportion of cases.
The convergence of these two forms creates a third, compound risk: companies that use AI washing both to inflate perceived capability and to obscure underlying financial deterioration. It is this compound form that is proving most fertile ground for securities class actions.
The Litigation Landscape
AI-related securities filings tracked 12 cases in the first half of 2025 alone – on pace to far exceed the 2024 annual total of 15. The Disclosure Dollar Loss Index reached $403 billion in H1 2025, a 56% increase from the prior half-year and the highest semi-annual total since the record set in 2022. The Maximum Dollar Loss Index reached $1,851 billion, marking the eighth consecutive semi-annual period above the historical average.
The target profile is broadening. While OpenAI, Google, and Meta remain frequent defendants – with ChatGPT cited most prominently in litigation – risk is spreading to smaller startups and to non-technology companies that have adopted AI tools without adequate disclosure governance. The plaintiff’s bar is now pursuing companies not only for what their technology is, but for how they characterize its financial impact.
Notable Cases Defining the Frontier
- Oracle Corp. – Alleged overstatement of how quickly AI capex would drive revenue; executives allegedly sold $1.87B in stock at inflated prices (Class action pending, U.S. District Court, Delaware)
- C3.ai, Inc. – Misrepresented revenue outlook; shares dropped over 25% after guidance was reduced (Ongoing securities litigation)
- Presto Automation – SEC charges for falsely claiming AI processed restaurant orders handled by humans (Settled January 2025)
- Innodata Inc. – Marketed as AI-powered business while relying on manual offshore labor (Active litigation)
- Nate, Inc. – Fraudulently raised $42M+ claiming app used AI; human contractors completed purchases (SEC enforcement action filed)
- Air AI – Marketed “conversational AI” to replace human service agents – $18M judgment (partially suspended)
- Click Profit – Promised passive income via AI-powered automation – Operators banned; combined judgments exceeding $20M
The Oracle case is particularly instructive for institutional investors and boards. The lawsuit does not allege that Oracle’s AI capabilities are fictitious – it alleges that management misled investors about the “timeline and magnitude” by which AI capital expenditure would convert to revenue, and that executives monetized $1.87 billion in equity while those disclosures were allegedly inflated. This represents a materially higher standard than prior AI washing cases: it is no longer sufficient to possess genuine AI capability. The way that capability is communicated to capital markets must be accurate, timely, and corroborated by disclosure controls.
The Regulatory Architecture
Launched in September 2024, the FTC’s Operation AI Comply initiative established the template for AI washing enforcement using existing consumer protection law – no new statute required. The agency brought five actions at launch; enforcement has continued under the new administration, which inherited the initiative and has kept it operational. The enforcement tools available to regulators are structural in their severity.
- Algorithmic Disgorgement: Companies may be forced to delete entire algorithms or AI models developed using unlawfully collected data, effectively erasing the R&D value embedded in those systems
- Permanent Industry Bans: Operators of deceptive AI schemes face lifetime exclusion from marketing or selling business opportunities
- Asset Forfeiture: Settlements require defendants to surrender cash, real estate, and personal property to fund consumer redress
These are not regulatory fines. They are structural penalties capable of destroying a business.
The Rytr Reversal: A Dual-Track Regulatory Posture
A significant interpretive development occurred in December 2025 when the FTC set aside its consent order against Rytr, a generative AI writing tool. Dissenting commissioners argued that the mere potential for a tool to be misused – for example, to generate fake reviews – does not justify enforcement without empirical proof that the tool actually caused consumer injury. The Trump administration’s AI Action Plan, which explicitly directs agencies to reduce enforcement where it “unduly burdens AI innovation,” provided the political context.
The result is a two-track regulatory posture that is now operative: fierce, structural enforcement against companies that deceive consumers about their own AI capabilities, combined with a materially more permissive stance toward platform and tool developers who are merely vulnerable to third-party misuse. For corporate counsel and compliance teams, the practical implication is clear – the regulatory exposure is concentrated in claims companies make about themselves, not in the downstream applications their tools enable.
The SEC’s enforcement trajectory is somewhat more uncertain following leadership changes, though the Nate, Inc. filing demonstrates continued willingness to pursue egregious product AI washing cases. Boards should note that the SEC’s 2026 Examination Priorities explicitly target AI-related disclosures.
The Agentic AI Amplifier
The next wave of AI washing is already forming around agentic AI. Gartner predicts that over 40% of agentic AI projects will be canceled by the end of 2027 due to escalating costs, unclear business value, and inadequate risk controls. The research firm found that only approximately 130 out of thousands of tools marketed as “AI agents” possess genuine agentic capabilities – the ability to take autonomous, multi-step action. The rest are repackaged chatbots, AI assistants, and robotic process automation tools operating under an “agent washing” rebrand.
This matters because companies are currently staking strategic narratives – and in some cases investor guidance – on agentic AI deployment timelines and productivity projections. If Gartner’s attrition projections materialize, a significant cohort of those companies will face the compound liability of having made aspirational claims that were then contradicted by operational abandonment. The securities bar will treat the gap between the claim and the cancellation as the evidential core of a misrepresentation case.
Systemic Implications for Deal-Making and Governance
M&A: The Shadow AI Problem
AI washing is fundamentally changing the due diligence calculus for private equity and corporate acquirers. The defining red flag in 2026 transactions is “shadow AI” – the condition in which a target company lacks centralized records of which AI systems are in use, what data has passed through them, and on what legal basis. ArmorCode’s 2026 State of AI Risk Management report found that while 86% of organizations claim to have a complete AI inventory, 59% simultaneously admit that shadow AI is present and ungoverned within their environments. Mimecast’s 2026 survey found that 80% of organizations are concerned about data leaking through generative AI tools, yet 60% lack any specific strategy to address it.
The due diligence implication is structural: where shadow AI is present, a record-based review of AI representations becomes impossible. Acquirers are forced to rely on management representations that are inherently harder to verify – and harder to warranty or indemnify against in transaction documents. PE firms conducting technology due diligence must now include AI inventory attestation, model provenance documentation, and data lineage records as standard requirements, alongside traditional IP and regulatory compliance reviews.
Legal Privilege and Sovereign AI
A February 2026 ruling in United States v. Heppner established that legal work product generated by AI tools without strict confidentiality protocols may not qualify for attorney-client privilege. The ruling is accelerating a structural shift in how law firms and corporate legal departments deploy AI – away from public platforms and toward private, sovereign AI deployments designed to prevent data leakage and preserve legal protections. For corporate boards, this ruling has a direct governance implication: AI tools used in legal strategy, board deliberation, or regulatory correspondence must now be assessed for privilege compatibility, not merely for performance.
D&O Insurance: The Coverage Gap
Standard Directors and Officers and Errors and Omissions insurance policies were not drafted with generative AI risks in mind. As claims involving hallucinated outputs, automated decision-making bias, and undisclosed AI usage have emerged, companies are encountering significant coverage disputes. The situation is evolving rapidly and not uniformly in policyholders’ favor.
For the first time in 2026, Verisk’s Insurance Services Office introduced an optional exclusion to CGL policies for loss “arising out of generative artificial intelligence”. Berkley Insurance Company has introduced an “Artificial Intelligence Exclusion (Absolute)” applicable to D&O and E&O policies. These are not edge cases – given the insurance industry’s reliance on ISO forms, policyholders should expect widespread adoption. The practical consequence is that AI washing claims – including those arising from board-level disclosure decisions – may not be covered at the next policy renewal cycle even if they would have been covered today.
Five Governance Imperatives for Boards
The consolidation of AI washing into a material legal and financial risk category requires boards to treat it as a standing governance obligation rather than a one-time compliance exercise. The following imperatives are grounded in the litigation and regulatory record reviewed above.
1. Establish an AI Claims Register
Every public-facing claim about AI capability – in investor presentations, earnings calls, product marketing, regulatory filings, and social media – should be registered, dated, attributed to a responsible executive, and supported by documented technical evidence. The Oracle case demonstrates that the most dangerous claims are not fabrications but extrapolations: accurate capability claims attached to inaccurate timelines or financial projections.
2. Conduct an AI Inventory Audit
Given that 59% of organizations have ungoverned shadow AI operating inside their technology stacks, boards cannot rely on management representations alone. An independent AI inventory audit – cataloging all tools, models, data flows, and third-party dependencies – should be a precondition for any AI-related disclosure and a standard component of M&A due diligence.
3. Separate AI Rationale from Workforce Decisions
The 42% of HR managers who cited AI as cover for financially-driven layoffs created direct securities fraud exposure for their employers. Boards must ensure that workforce communication is grounded in disclosed financial rationale, not AI narratives constructed after the fact. The reputational and legal exposure from operational AI washing is now equivalent to that of product AI washing.
4. Stress-Test Insurance Coverage
Policy renewals in 2026 require active negotiation to eliminate or narrow AI-specific exclusions that insurers are now introducing. Boards should require their risk committees to report on the specific coverage position for AI washing claims, including securities fraud arising from AI disclosure, before the next renewal cycle.
5. Apply Legal Privilege Protocols to AI-Assisted Deliberation
The Heppner ruling means boards using AI tools in legal, strategic, or governance contexts must ensure those deployments meet privilege requirements. Sovereign, private AI deployments – where data does not leave the organization’s controlled environment – are now a legal necessity rather than a preference for security-conscious organizations.
Outlook
The trajectory is unambiguous. AI washing litigation will continue to grow in volume and sophistication as the plaintiff’s bar develops case law, expert witnesses, and evidential frameworks tailored to AI disclosure disputes. The agentic AI cycle will generate its own wave of class actions as the gap between marketed capability and operational delivery becomes measurable. Regulatory enforcement will remain active at the FTC for outright consumer deception, while the SEC’s posture on AI disclosure accuracy – still directionally hawkish despite political headwinds – will be the dominant variable for public companies.
The companies that navigate this environment successfully will be those that treat AI transparency not as a communications problem but as an operational discipline – with the same rigor, documentation, and governance infrastructure applied to financial disclosure. The ones that do not will find themselves litigated into that discipline through a far more costly process.
Disclaimer
This article has been prepared by AIMG (AI Markets Group) for general information and discussion purposes only and does not constitute legal, investment, tax, or other professional advice. Readers should not act or refrain from acting on the basis of this material without obtaining advice from appropriate professional advisers. While reasonable care has been taken to ensure the accuracy of the information contained herein, AIMG makes no representation or warranty, express or implied, as to its completeness, reliability, or accuracy and accepts no responsibility or liability for any losses or damages arising directly or indirectly from any use of, or reliance on, this publication. Any opinions expressed are subject to change without notice.