AI-Washing Claims

AI-washing is the overstatement or misstatement of artificial intelligence capabilities, or of AI’s contribution to revenue, in public filings and investor presentations. In 2026 it became one of the main sources of securities class actions in the United States, and underwriters of directors and officers (D&O) insurance are re-examining every technology company that tells its story through AI.

For an Israeli company listed on Nasdaq, or preparing an IPO, the D&O policy is the first line of defence against such a claim. The question is whether the existing policy still responds to a claim built entirely on AI statements, at a time when part of the market is adding AI exclusions.

Key points

An AI-washing claim is a conventional securities claim. The allegation is that the company made a misleading statement about AI and the share price fell when the truth emerged. In the US it is usually brought under Section 10(b) of the Exchange Act and Rule 10b-5, and for a recently listed company also under Section 11 of the Securities Act.

The 2026 numbers stand out. Industry data published in July 2026 points to 121 US securities class action filings in the first half of the year, up 30% on the previous half. Fifteen were AI-related, and they accounted for roughly 73% of total Disclosure Dollar Loss.

The underwriting table has changed. Underwriters now ask how AI generates revenue and what it costs. Some insurers have introduced broad AI exclusions in management liability policies, while others offer endorsements that affirm cover.

Some things insurance cannot buy. Israel’s Companies Law prohibits insuring an officer against a fine, civil fine or monetary sanction, and the Securities Law prohibits insuring an administrative enforcement sanction. Defence costs and payments to injured investors may be insurable, subject to the articles and the wording.

How an AI statement becomes a class action

An AI-washing claim is born when the gap between what the company said about AI and what later emerged translates into a share price drop. Plaintiffs do not need to prove that the AI does not work. It is enough to allege that the company presented it as better or more profitable than it was, and that investors relied on that.

The suits filed in the US since 2024 follow four recurring patterns. The first is overstated product capability, including presenting third-party technology as in-house development. The second is growth attributed to AI, sometimes alongside a multi-year revenue target that was dropped months later. The third is relatively new: an allegation that the company understated the development and infrastructure cost of its AI products. The fourth is concealing the competitive threat from rivals’ AI tools. The corrective disclosure usually arrives with a weak quarterly report, and the suit follows within weeks.

The US Securities and Exchange Commission (SEC) opened this line before the class action bar did. In March 2024 it settled proceedings against two investment advisers, and in January 2025 it brought its first action of this kind against a public company. In April 2025, parallel civil and criminal cases were brought against a startup founder who raised more than $40 million from private investors on the claim that his product was AI-driven, when human workers did most of the work. The SEC’s 2026 examination priorities still list the accuracy of AI capability claims, but the centre of gravity has moved to shareholder litigation.

For an AI startup raising capital, the case against the founder matters more than any class action: exposure starts with the investor deck, long before a listing.

What changed in 2026: the data and the Israeli companies

In 2026, AI claims became the heaviest component of the US securities class action market. They are a minority by count, they concentrate most of the financial exposure, and they have already reached Israeli companies.

AI filings were about 13% of filings in the first half of 2026, yet about 73% of both Disclosure Dollar Loss and Maximum Dollar Loss. Underwriters use these two measures to size a potential claim. Filings against non-US issuers listed in the US also reached their fastest pace since 2020.

In 2026, class actions were filed in the US against at least two Nasdaq-listed Israeli technology companies over AI-related statements. In one, the company allegedly reaffirmed a revenue target that rested on AI investment and was dropped months later. In the other, the company allegedly presented its flagship AI product as more competitive than it was and understated its cost. A 2024 suit against a third Israeli company, over how it described its AI capabilities around its IPO, was dismissed in full in September 2026, after more than two years in which defence costs ran against the limit.

How a D&O policy is meant to respond: the three sides

A public company D&O policy is usually built in three parts, and an AI-washing claim may trigger all three at once. The question is how they share one limit, and what happens when the wording contains an AI-related exclusion.

Side A protects directors and officers when the company cannot or may not indemnify them, for example in insolvency. Side B reimburses the company for indemnity it has paid. Side C, which for a public company is usually limited to securities claims, covers the company itself as a defendant. In a US class action the company and its officers are sued together, so entity cover reduces allocation disputes, as explained on the page on D&O for publicly traded companies. The price is a shared limit the company can erode on its own, which is why many companies add a dedicated Side A layer.

D&O responds to claims by shareholders and regulators against management and the company. A claim by a customer harmed by an AI product belongs to technology errors and omissions (Tech E&O) cover, which responds to product or service failure and stops short of securities claims. The split between the lines is set out in the coverage map for Tech E&O, cyber and D&O.

AI exclusions and underwriting questions: what is happening in the wordings

Until 2025 most D&O policies were silent on AI. Over the past year the market has split: some insurers have added “absolute” exclusions, some offer explicit endorsements, and most of the change happens quietly through definitions and proposal forms.

An absolute AI exclusion is usually drafted to exclude any claim “arising out of” the use of AI. For a company whose business story is AI, that wording may empty securities cover of content, because almost every public statement touches AI. The difference between “arising out of” and “based upon” may decide an entire claim. This is the management liability version of the gray areas in AI failure cover.

The 2026 proposal forms ask AI companies to explain how AI generates revenue and how resilient the business model is to disruption. Public company D&O was relatively stable at 2026 renewals, but underwriter surveys from September 2026 expect rate increases in 2027 and higher retentions for newly public companies.

The broker (LAMDA Broking) and the company’s counsel read the AI definitions in the renewal draft against the business description in public filings, so that the exclusion does not bite into the story the company tells investors.

The Israeli side: what may be insured and where shareholders sue

Israeli law determines what may be insured and the framework for shareholder claims in Israel. For a dual-listed company it operates alongside US law.

Sections 258 to 263 of the Companies Law allow a company to insure an officer for breach of the duty of care, breach of the duty of loyalty in good faith, and monetary liability to a third party. An advance indemnity undertaking is limited to events the board foresaw and to an amount it set as reasonable. The law voids insurance and indemnity for an intentional or reckless breach of care, for an act intended to yield unlawful personal profit, and for a fine, civil fine, monetary sanction or payment in lieu. Section 56H of the Securities Law prohibits insuring an administrative enforcement sanction, and allows insurance of a payment to the injured party and of proceeding costs, if the articles permit.

As of the date of writing, the Israel Securities Authority has not published issuer-specific guidance on AI disclosure. An interministerial report on AI in the financial sector, published in December 2025 under the joint lead of the Authority and the Ministry of Justice, warned against AI-washing and recommended handling it with existing tools. Existing disclosure rules apply to reporting companies in full, including the safe harbour for properly identified forward-looking information.

The Class Actions Law permits class actions on securities grounds, heard in the Economic Department of the Tel Aviv District Court. For a dual-listed company, which law governs a buyer of shares on Nasdaq has been addressed by Israeli courts, and the answer depends on the facts and requires individual advice. The business judgment rule protects decisions made in good faith, without conflict of interest and on an informed basis. A board that never asked to see the technical basis for the company’s AI statements will struggle to show it decided on an informed basis.

Europe and Asia: the AI Act as a source of disclosable facts

In Europe the risk comes less from class actions and more from regulation that creates facts to disclose. For a company selling AI products in Europe, the EU AI Act adds obligations whose breach may become a disclosure item, and from there a cause of action in the US.

The transparency obligations in Article 50 apply from 2 August 2026: people must be told they are interacting with an AI system, and synthetic content must be marked in a machine-readable format. Generative systems already on the market have until 2 December 2026 to comply with the marking duty. Regulation (EU) 2026/1744, the Digital Omnibus on AI, entered into force on 27 July 2026 and moved high-risk obligations to 2 December 2027, and to 2 August 2028 for systems embedded in regulated products. Supporting standards are still being prepared. Fines for breaching the transparency duties reach EUR 15 million or 3% of worldwide turnover. D&O policies usually exclude fines or cover them only where the law allows.

Companies also listed in Europe are subject to the Market Abuse Regulation, and the European securities regulator has not yet issued guidance specific to issuer AI-washing.

In Asia-Pacific, Australia’s securities regulator warned companies and directors in 2024 that it was watching for AI-washing. In most Asian financial centres, regulators have focused on supervised financial firms rather than issuers. A company active in the region should check the policy’s territorial limits.

Transitions between policies: where continuity breaks

Most coverage gaps in AI claims open in the transition between policies: at a fundraise, at an IPO, on a change of insurer or on a sale. For a claim over an old statement, which policy was in force matters as much as the wording.

An AI startup raising capital is exposed from the investor deck onwards. Private company D&O usually covers investor claims in a private round and tends to exclude public offerings. Its conduct exclusion should require final adjudication, so that defence costs are advanced until the case is decided.

On the day of a Nasdaq IPO, the private policy is replaced by a public company policy. The description of AI capabilities in the prospectus exposes the company to a Section 11 claim, in which the plaintiff need not prove intent. The private policy goes into run-off for pre-IPO acts.

On a merger or acquisition, the change in control clause puts the policy into run-off, and a six to seven year discovery period is common. On a change of insurer, a short-seller report or regulatory inquiry received before renewal should be notified to the expiring policy as a circumstance, and the continuity date should be preserved in the new policy.

Common mistakes

• “This is a risk for AI companies only”: any company that describes its use of AI in a filing or a deck is exposed to an allegation that the description was overstated.
• “An AI exclusion only concerns product losses”: absolute exclusions have entered management liability policies, and some reach securities claims.
• “The limit is reserved for the directors”: when the company is itself sued and covered, the shared limit erodes fast.
• “An AI Act fine will be covered”: fines and sanctions usually remain outside cover. Defence costs in an investigation are a separate question.
• “Changing insurer is a price decision”: switching without preserving the continuity date may leave a claim over an old statement uncovered.

Frequently Asked Questions

Our company does not develop AI, it only uses it in its product. Are we still exposed to an AI-washing claim?

Yes. Exposure follows what the company says about AI. A company that attributes improved profitability to AI, and later reports results that do not match the description, may face the same allegation. The 2026 filings include infrastructure and data centre companies alongside model developers.

Does a D&O policy cover a class action over overstated AI capabilities?

A public company D&O policy is designed to respond to securities claims, and an AI-washing claim is one. Cover depends on the wording, above all on whether an AI exclusion has been added and how the conduct exclusion is drafted. Subject to the wording, defence costs are funded even when the claim is eventually dismissed.

The insurer added an AI exclusion at renewal. What should we check?

Check its scope. Wording based on claims “arising out of” the use of AI may reach almost any statement a technology company makes. Ask for a carve-back for securities claims and for Side A. If the insurer refuses, consider an alternative market.

Is a fine under the EU AI Act or a sanction by the Israel Securities Authority insurable?

Generally not. Israel’s Companies Law prohibits insuring an officer against a fine, civil fine or monetary sanction, and the Securities Law prohibits insuring an administrative enforcement sanction. A payment to the injured party and defence costs may be insurable, subject to the company’s articles and the policy wording.

We are an AI startup ahead of a funding round. Do we need D&O now?

Institutional investors usually require D&O as a condition of taking a board seat. In 2025 a founder who raised money from private investors on a false description of AI capabilities was indicted in the US. A private company policy usually covers investor claims in a private round, subject to exclusions.

We are heading for a Nasdaq IPO. What changes in the policy on IPO day?

The private policy goes into run-off, and a public company policy with entity cover for securities claims is bought. The description of AI capabilities in the prospectus exposes the company to a Section 11 claim. Settle the policy structure months before the offering.

What drives the cost of D&O insurance for an AI company?

Price follows market capitalisation and share volatility, the extent of US activity, claims history and the quality of disclosure. For AI companies underwriters also look at how AI is described in filings and at the ratio of AI spending to AI revenue. Indicative pricing is never binding and depends on the full underwriting picture.

An AI-washing claim rests on what management told investors, so the D&O policy should be read against those same disclosures, line by line, before the next renewal.

The above is general information only and does not constitute insurance, legal or other professional advice. Coverage terms, exclusions and obligations are determined by the specific policy wording and subject to applicable law. Each case should be reviewed on its merits against the policy wording and with a qualified adviser.