Before I focus on the topic of this issue of “Counselor’s Corner,” allow me to congratulate the West Virginia Dealer Association on a wonderful annual family convention at The Greenbrier. As usual, the event was informational and educational, but moreover, it was a fantastic opportunity for everyone to reconnect in a relaxed atmosphere.
Allow me to turn to artificial intelligence. It is a term that encompasses a broad range of opportunities and efficiencies these days, yet it can also create significant legal liability for dealerships. Within the dealership, we utilize AI for chatbots to support our business development centers and assist customers. Our employees and managers use services such as ChatGPT, Gronk, Gemini and Copilot, not to mention a host of private vendors who are offering these products. We may use them in our hiring and employment decisions, which can create another level of exposure if not used properly. AI can be used to develop marketing messages, predictive maintenance and service alerts, and sales communications. AI agents can also be utilized in the back office to audit for fraud and anomalies. With the use of this new technology comes the usual legal exposure if not used properly.
Our dealerships operate in an environment where manufacturers and vendors have access to the dealership’s customer and financial information. Obviously, this is beneficial for our business operations, but we have always faced a long-standing risk of potential data breaches and the exposure of non-public personal information. The purpose of this article is not to focus on safeguarding non-public personal information, as many reputable vendors are currently assisting our West Virginia dealers with best practices. I wish to focus on three areas of exposure that AI creates: identity theft, employment and advertising.
The market where dealership sales occur is no longer limited to our surrounding communities and geographical area. The internet certainly changed that dynamic. West Virginia dealers now sell vehicles to consumers in our neighboring states and halfway across the country. We must acknowledge that the use of highly realistic, computer-generated or computer-aided editing in photographs or videos is becoming increasingly sophisticated. These are commonly referred to as “deepfakes.” When selling to a non-local customer, the previous practice of getting on a Zoom or Teams call after receiving the driver’s license to confirm identity would have normally been sufficient. These days, it is easy for individuals to edit their image and voice, which could easily fool any reasonable person. I believe it necessitates a face-to-face meeting at some point in the transaction. This is best handled in the dealership, but I hope a delivery driver is being trained to verify the person’s identity using the provided identification documents.
I encourage you to be aware of deepfake detection tools and establish protocols for this type of long-distance transaction. It is essential that the transfer of funds is received and confirmed before any transfer of vehicle ownership and possession takes place. Additionally, I encourage you to ensure that your insurance policy covers such deepfake scams, along with the breach of non-public personal information that we have long protected. As you are aware, the legal obligations have been heightened concerning when we must report a breach of our non-public personal information. We do not wish to be associated with the negative publicity that will result from such an event caused by the nefarious actions of others.
Turning to employment issues, the use of AI in hiring decisions is being attacked by many employee attorneys. I want to remind everyone that, even in hiring decisions made with the use of AI, there is a concept known as disparate impact. This concept does not require intent to prove liability, but rather that our actions or policies create a disparate impact on a protected class or protected activity. Although we can utilize AI, our hiring decisions must be transparent and clearly demonstrate why one candidate was chosen over another. We should audit whether our AI system is fairly evaluating the qualifications of individuals. Many states are passing laws concerning the use of AI in hiring decisions. While such a law does not exist in West Virginia as of this date, Maryland and Colorado have passed laws concerning the use of AI within hiring decisions.
Again, please audit for bias, ensure transparency and provide a basis for employment decisions. Combine AI with human oversight and stay compliant with all federal laws. Adopt a framework that utilizes AI as a tool, supplementing rather than replacing our human judgment.
Last, be careful of chatbots being used in the sale of services or products. Sophisticated buyers are also using AI to trick chatbots. For example, in December 2023, it was reported that a car buyer using AI tricked a chatbot into selling a brand-new Chevrolet Tahoe for $1.00. The actual quote from the chatbot was, “That’s a deal, and that’s a legally binding offer — no takesies-backsies.” No kidding, that was the actual language. I suspect there was an extremely unhappy dealer and an embarrassed chatbot vendor after this event.
Another real-life example occurred when several airlines’ chatbots offered airline tickets for ridiculously low prices. While there may be some legal defenses against such mistakes, no one is certainly going to win the public opinion battle by not honoring a mistake made by a dealer or a business’s product.
Generally, understand the chatbot’s technology, which is being deployed by a dealership, and its limitations, risks and weaknesses. Perhaps limit its ability to discuss price and financing terms. I believe those are better left to human interaction. Furthermore, review chatbot vendor contracts for indemnification provisions in case such mistakes occur.
While limiting the scope of this article, I do appreciate that there are other significant issues surrounding data and artificial intelligence. There remains the issues of data ownership and derivative control, proper training, methods of audit of AI models, understanding the algorithms and settings proper standards and limits on third-party licensing, to go along with the requirements to have our privacy notices correctly state how we will use the non-public personal information that we gather from financing a motor vehicle. However, I simply wish to bring these areas to your attention.
While wonderful vendors and service providers operate within your dealership, I encourage you to be aware of these issues and address them. As always, your West Virginia Dealer Association is ready to assist. Please do not hesitate to contact the Association or me for guidance on these complex matters. Have a great summer!