When AI Marketing Gets Ahead of AI Engineering

Earlier in May, Apple agreed to pay $250 million to settle a consumer class action about how it marketed Apple Intelligence and the upgraded Siri features for the iPhone 16. The claims were never ultimately proven in court, and Apple did not admit any wrongdoing. But the theory of the case—that consumers paid a premium for AI features that were marketed as ready and weren’t—is similar to what the SEC and FTC have been flagging for two years. Companies may face enforcement action and private litigation risk if they exaggerate AI capabilities that don’t yet exist, overstate what the AI actually does, or attribute AI to product features that are not materially driven by AI.

Landsheft v. Apple Inc.

Landsheft v. Apple Inc., No. 5:25-cv-02668 (N.D. Cal.) is a case about timing. Like many makers of connected devices, Apple intended to roll out a series of AI features over time after the launch of the product. Here, the plaintiffs argued that the marketing that ran ahead of delivery falsely advertised those Apple Intelligence features that were not yet available to customers. The Apple settlement covers all iPhone 16 models and certain iPhone 15 models purchased between June 10, 2024 and March 29, 2025, roughly 37 million devices. 

The motion for preliminary approval, filed May 5, is worth reading closely. Plaintiffs’ counsel acknowledged this was uncharted territory, noting that they were “not aware of any prior cases involving a high-profile marketing campaign involving alleged false advertising about AI technology” at this scale, and flagged “legal issues of first impression” around whether consumers can demonstrate reliance on AI-feature representations and whether damages can be isolated from a multi-feature product. There are, they noted, “few trial court opinions and virtually no appellate court opinions involving AI products and services.” 

Apple previewed defenses that may be relevant in other AI-feature cases across the market: it disclosed upfront that features would roll out over time; it did, in fact, ship certain advertised features that were available at launch; and consumers buy iPhones for plenty of reasons besides Siri. Apple also argued that any alleged damages disappear once advertised features ship, pointing to Waller v. Hewlett-Packard Co., 295 F.R.D. 472 (S.D. Cal. 2013), where a software update mooted the claim. That “delivery cures deception” argument was not tested here, but could provide a powerful defense in future cases.

Plaintiffs, meanwhile, ran a consumer survey that purported to identify a 5.5% price premium attributable to the “Enhanced Siri” marketing. Applied across 37 million devices at an average price of roughly $1,110, that yielded about $2.26 billion in estimated class-wide alleged damages. The $250 million settlement is around only 11% of that ceiling, meaning companies making bullish AI claims could face significant financial exposure. 

Claims at issue in Landsheft were prefaced by NAD Challenge

While Apple and the plaintiffs in Landsheft noted that there are few federal or state consumer class actions on the issue, there are numerous NAD cases and regulatory actions targeting so-called “AI-washing,” which have no doubt garnered the attention of the plaintiffs’ bar. 

In early 2025, the NAD reviewed Apple’s Apple Intelligence claims and recommended that Apple modify or discontinue several of them in Apple, Inc. (Apple Intelligence), Report #7410, NAD/CARU Case Reports (April 2025). NAD alleged that Apple’s advertising featured an “Available Now” banner, without qualification, sitting above feature descriptions for things like Priority Notifications, Image Playground, Genmoji, ChatGPT integration and new Siri capabilities—features Apple later rolled out in staggered updates over six months. The NAD said that footnote disclosures didn’t adequately qualify the “Available Now” claim, and as a result of the NAD’s recommendations, Apple pulled the challenged demo video down entirely.

Although the original complaint in Landsheft was filed a month prior to the NAD’s press release on the inquiry, it’s worth noting that plaintiffs’ counsel quoted the NAD’s findings directly in their amended complaint. So, while NAD decisions are voluntary and non-binding, they are now showing up in federal complaints as alleged evidence of deceptiveness and materiality. That should change the calculus for any company deciding how seriously to take an NAD inquiry about its AI claims. 

Wider Context

Regulators were already moving before the Apple settlement. The SEC charged Delphia and Global Predictions in March 2024 for misrepresenting their use of AI and obtained $400,000 in penalties. SEC Enforcement Director Gurbir Grewal put it bluntly, “If you are rushing to make claims about using AI … stop. Take a step back, and ask yourselves, ‘Do these representations accurately reflect what we are doing or are they simply aspirational?’” The FTC launched “Operation AI Comply” that September, with various enforcement actions targeting fake review tools, “AI Lawyer” claims, and AI-powered income schemes. 

The FTC’s enforcement posture has not softened with the change in leadership in 2024. In a September 2025 keynote, Chairman Andrew Ferguson stated that the agency “has repeatedly deployed its consumer-protection authority to protect consumers from false and misleading claims about AI products’ capabilities.” Bureau of Consumer Protection Director Christopher Mufarrige confirmed at the NAD conference that same month that the Commission is “targeting fraudsters and other unscrupulous actors who make false and misleading claims related to their use of AI.” Since Operation AI Comply launched in late 2024, the FTC has brought or resolved more than a dozen AI-related enforcement actions. The underlying legal theory is not new—these are Section 5 false advertising cases—but an “AI” label on the product or service is what draws regulatory attention, and the pace is accelerating.

• NGL Labs LLC – in a complaint filed with the Los Angeles County District Attorney’s Office, the FTC claimed that NGL falsely claimed that its automated moderation filters, filtered out harmful content from its anonymous messaging app, which was marketed to children. In a proposed order in July 2024, the NGL agreed to implement measures to prevent users under 18 from accessing the app, pay $4.5M to the FTC and a $500K civil penalty to Los Angeles County.
• Evolv Technologies – the FTC claimed that Evolv made unsupported claims by stating that its AI-powered security screening system was capable of detecting weapons, while ignoring harmless personal items such as laptops, lunchboxes, etc. The company also failed to disclose its involvement in the design of the third-party testing report and that it removed negative information about the tool’s effectiveness from such report. In its settlement order in November 2024, the FTC ordered the company to offer its clients an option to cancel their contracts. Following the FTC probe, the SEC opened an investigation and its shareholders filed a complaint.
• IntelliVision – in January 2025, the FTC finalized an order against IntelliVision, claiming that IntelliVision misrepresented its AI-based facial recognition software by stating that it was free of racial and gender bias, that it had the highest accuracy rates on the market, and that its anti-spoofing technology could identify a photo or video image. Testing showed the software had similar issues with race and gender as other similar products in the market, showing particular difficulty with certain groups.
• DoNotPay Inc. – the company claimed that its AI chatbot was the “world’s first robot lawyer” and that it could operate at the level of a human lawyer when generating legal documents and giving advice. However, it did not hire or retain lawyers to test the quality or accuracy of the services. In its final order in February 2025, the FTC ordered DoNotPay Inc. to pay $193K and to notify consumers who subscribed to the service about the settlement.
• accessiBe – the FTC claimed that accessiBe misrepresented that its AI-powered tool could make websites compliant with international accessibility guidelines. The April 2025 settlement agreement for these claims included a $1M payment to the FTC.
• Workado – the FTC claimed that Workado made deceptive claims about its AI-powered text-detector software by stating that it could classify text as either human-created or AI-generated with over 98% accuracy, but testing showed it only had 53% accuracy. The FTC’s proposed order in April 2025 required Workado to stop advertising the accuracy of its AI detection products without competent and reliable evidence.
• Growth Cave, LLC – in January 2026, the FTC resolved its case against Growth Cave, which had promised that its “AI software” GrowthBox would “automate nearly 100% of the process” of running an online education course. The technology actually required users to do most of the work manually. The settlement bars defendants from misrepresenting that a product uses AI to enhance profitability or efficiency, covering both claiming AI when there is none and overstating what real AI actually does, and requires one of the co-CEOs to liquidate personal assets to help fund the nearly-$50 million judgment. 
• Air AI Technologies, Inc. – the FTC sued Air AI for claiming its product could autonomously replace human customer-service representatives and generate substantial earnings for small businesses. The FTC alleged that consumers lost as much as $250,000. The March 2026 settlement imposed an $18 million judgment and a permanent ban on Air AI’s operators from marketing business opportunities, and may be the first consumer-protection action targeting deceptive claims about agentic AI.

While there are some signals of a pullback from the very aggressive stance former FTC Chairman Linna Khan’s Commission took in AI matters in December 2025, the FTC reopened and set aside its 2024 consent order against Rytr, an AI writing tool accused of facilitating fake reviews, finding the original complaint did not meet the agency’s standards. Ferguson’s FTC simultaneously issued 6(b) orders to seven companies offering AI companion chatbots, demanding information about their safety and data practices, and it has continued filing new cases into 2026. These actions illustrate a clear message being sent by regulators: companies that make specific, testable claims about what their AI can do—accuracy rates, automation levels, earnings potential—should expect those claims to be tested.

Takeaways

Gate the marketing behind the engineering. The Landsheft case turns on allegations that Apple’s marketing got ahead of its delivery schedule. Similar allegations show up across FTC enforcement actions, with companies allegedly claiming accuracy rates, automation levels or capabilities that the product allegedly cannot deliver at the time of launch. Product and legal need to sign off that a feature works as described before anyone markets it or qualify the marketing to clarify when any features will be available.

Damages defenses may be available but remain untested. Apple cited Waller v. Hewlett-Packard Co., 295 F.R.D. 472 (S.D. Cal. 2013), for the proposition that damages are not inflicted if a software update delivers the advertised feature. The settlement means no court actually ruled on it. Clients who plan to ship AI features post-launch should assume, absent case law on the topic, that there remains potential exposure in the gap between the ad and the update. 

Take NAD seriously. Every admission, modification and voluntary pull-down in a NAD proceeding is now potential exhibit material in a federal complaint. Coordinate the NAD response with litigation counsel from the start.

Fix the headline, not the footnote. NAD said Apple’s small-print qualifications couldn’t save the “Available Now” banner, and the FTC’s recent cases reflect the same principle. If a product requires significant manual effort, an “AI-powered” headline won’t be saved by easily-missed disclaimers. If a feature is staged, limited, or still requires human involvement, that limitation may counsel against aggressive “AI-powered” claims, or at a minimum, clarification about the claim belongs in the headline, the product listing and the pitch deck. 

Demo videos are claims. Apple pulled its NAD-challenged ad after the Siri features it depicted were delayed. Google unlisted a Gemini demo after NAD inquired. Both cases illustrate that claimants could argue that videos showing AI tool capabilities imply current use cases. Curated prompts, cherry-picked outputs, edited response times—all of that needs legal review before the video goes live.