NAD Decision Provides Guidance On Substantiating AI Claims

At last year’s ANA Masters of Advertising Law Conference, Phyllis Marcus, Vice President of NAD, stated that NAD would closely scrutinize claims related to AI and hinted that new decisions were on the way. We wrote about one of those decisions in January. Not long after, NAD announced another decision that involves AI claims. (Coincidentally, both cases involve baby monitors.) The second decision, in particular, outlines a helpful framework for considering AI-claims.

Every parent of a newborn struggles to figure out what their child is trying to communicate. (Many parents of teens struggle with this, too, but that’s for another post.) Dorel Juvenile Group claims it can help parents know what their babies are saying. The company advertises that the CryAssist technology included in some of its bassinets uses “AI to translate your little one’s cries, letting you know if they might be sleepy, fussy, gassy, agitated, or hungry.” 

NAD emphasized the importance of ensuring that AI claims are substantiated. In a key paragraph of the decision, NAD explains that “understanding the validity of claims about AI requires a focus on both the pre-deployment training data used to develop and teach the model, the testing and validation of the model to determine how well the model has learned and performs on new data, and verification that the specific AI model performs in the product being sold.” 

NAD walked through each of these areas. The evidence presented by Dorel showed the AI could distinguish between different types of cries with about 92% accuracy (consistent with existing research showing accuracy of between 89-94%). Importantly, Dorel was also able to demonstrate that the technology maintained this level of accuracy after it was integrated in its bassinets. Accordingly, NAD determined that the company’s claims were substantiated. 

As with the decision we discussed in January, CARU also considered whether Dorel’s data collection practices comply with COPPA and CARU’s Privacy Guidelines. CARU found that Dorel maintains reasonable security measures, limits data collection to information necessary to support the service, and that it doesn’t use children’s data for undisclosed secondary purposes. However, CARU recommended that Dorel improve its notice and consent procedures. 

As more companies start to integrate AI into their products and advertise the benefits of their technology, we expect to see an increase in the number of challenges involving those ads from multiple angles—including from NAD, competitors, and consumers in class actions. The type of substantiation a company will need to support its claims will depend on the context, but this decision provides a good starting point for thinking about that issue. 

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.