SEC and DOJ AI-washing actions have made AI-related misrepresentation claims a live antifraud pattern. For CFOs signing periodic reports that reference AI, the practical question is whether the disclosure-controls record supports the words used in filings and investor communications.
In brief: The SEC brought AI-washing settlements against investment advisers in March 2024, charged Joonko founder Ilit Raz in June 2024, and brought a parallel SEC/DOJ action against Nate founder Albert Saniger in April 2025. Cornerstone Research reported 16 AI-related securities class action filings in 2025. An audit-trailed, four-eyes approved record of what “AI-powered” means inside the company is one artifact that can support disclosure-control documentation.
The SEC's March 2024 Delphia and Global Predictions settlements treated allegedly false AI use claims as securities-law matters for investment advisers. The June 2024 Joonko/Raz action and the April 2025 Saniger action moved the fact pattern into named-executive startup fundraising claims.
The litigation pipeline is also visible. Cornerstone Research reported 16 AI-related securities class action filings in 2025, with 12 filed in the first half and four in the second half. AI-themed complaints now feed into the broader securities litigation docket; the average AI-specific settlement figure remains sparse and should not be treated as mature actuarial data.
For CFOs, the actionable point is not a new standalone AI disclosure rule. It is the ordinary public-company controls question: if a periodic report, earnings script, prospectus, investor deck, or board pack uses AI capability language, who verified that the words are accurate and not misleading?
The clearest illustration is SEC v. Saniger.
In April 2025, the SEC and DOJ filed parallel civil and criminal actions against Albert Saniger, the former CEO of Nate, Inc. The agencies alleged that Saniger raised more than $42 million from investors by falsely claiming his mobile shopping app used AI to complete purchases automatically, when transactions were allegedly processed manually by workers. Sources: SEC v. Saniger litigation release and DOJ Saniger release.
The pattern is not obscure. A company describes a capability as “AI” in investor materials; the underlying process is allegedly materially different; the gap becomes an antifraud problem. The Saniger charges are allegations, not findings, but they show how prosecutors and regulators can frame the gap.
Nate was a private fundraise, so it is not a Reg S-K periodic-reporting case. The lesson for public companies is parallel rather than identical: material AI statements in a 10-K, 10-Q, prospectus, earnings script, or investor deck need support from disclosure controls, factual review, and legal review. If a 10-K says the product is AI-powered, the internal definition of “AI-powered” has to support the claim.
The CFO is a named signatory on Form 10-K and 10-Q periodic reports and signs the SOX 302/906 certifications that accompany those periodic reports. S-1 registration statements have their own signature and liability framework; they are not SOX 302/906 certifications. In both settings, unsupported AI capability language can become a controls and verification problem.
Two practical questions drive the CFO review: did the company have reason to know the AI statement could be misleading, and what process existed to prevent that outcome. The second is where a CFO either has a paper trail or does not.
The risk is also a board and D&O renewal issue. AI-themed complaints enter the broader securities litigation and insurance pipeline. Audit committees ask the follow-up question: show us the process.
Under 15 U.S.C. § 7241 (SOX 302), the CFO personally certifies that the periodic report does not contain any untrue statement of material fact and that disclosure controls are designed to ensure material information is made known to the certifying officers. Under 18 U.S.C. § 1350 (SOX 906), the CFO certifies that the periodic report fully complies with Exchange Act requirements and fairly presents the financial condition. A knowing 906 violation carries a fine up to $1 million and up to ten years' imprisonment; a willful violation raises the ceiling to $5 million and up to twenty years. An unsupported material “AI-powered” claim in a 10-K is a cleaner SOX 302 disclosure-controls problem; SOX 906 exposure is narrower and depends on whether the false certification affects Exchange Act compliance or fair presentation of financial condition and results. The SEC's March 2024 settlements with investment advisers Delphia and Global Predictions, its June 2024 charges against Joonko founder Ilit Raz, and the Saniger/Nate case show related AI-washing and AI-adjacent antifraud patterns; they do not all allege the same type of AI capability misstatement.
Late-stage private and newly-public companies raise on AI capability claims. When those claims are precise, internally consistent, and supported by an audit-trailed definitional record, finance and legal teams can reduce terminology reconciliation work during diligence and S-1 drafting. For finance teams preparing public filings that reference AI, a versioned glossary creates an internal record of what each term — “AI-powered,” “machine learning,” “automated,” “autonomous,” “large language model” — means in the company's usage.
Compliance Glossary is not a disclosure-controls program. It is one artifact inside one. The narrow job it does is to show that the company took defined, timestamped actions to prevent misleading AI language from entering filings, earnings scripts, investor decks, or diligence materials.
We do not claim this removes AI-washing liability. Nothing a vendor sells does that. What it produces is the defensible record regulators and plaintiffs have asked for in analogous cyber-disclosure enforcement.
For current pricing, see the Atlassian Marketplace.
This article is informational and is not legal advice. Consult qualified securities counsel for disclosure decisions specific to your facts.
AI washing is the practice of overstating or misrepresenting a company's use or capability of artificial intelligence in investor, public-company, or marketing communications. SEC and DOJ actions show several patterns: allegedly false AI-use claims by advisers, allegedly false AI automation claims in Saniger/Nate, and broader startup-fraud allegations in Raz/Joonko where AI and automation were part of the investor story. This article is informational and is not legal advice.
The Nate matter involved a private company, so it is not a public-company periodic-reporting case. It is still relevant to CFOs because the antifraud theory turns on material investor statements about AI capability. Public-company CFOs separately sign Forms 10-K and 10-Q and SOX 302/906 certifications, so AI statements in periodic reports need disclosure-control support.
Regulators and plaintiffs ask what actions the officer took to prevent misleading disclosures. A four-eyes approved, version-controlled glossary of terms like AI-powered, machine learning, automated, autonomous, and large language model, with named approvers and timestamps, is a concrete artifact that may support a reasonable-care argument. Whether scienter can be challenged is a question for securities counsel on the specific facts.
Last verified: 2026-06-22. Sources checked: SEC v. Saniger, DOJ Saniger release, SEC Joonko/Raz release, SEC Delphia/Global Predictions release, Cornerstone 2025 filings report, 15 U.S.C. § 7241, and 18 U.S.C. § 1350.
For current pricing, see the Atlassian Marketplace.
Evaluate in Confluence Read the Security Whitepaper