Unveiling Truths, Connecting Communities

Unveiling Truths, Connecting Communities

Close this search box.

AI Chatbot Backfires On Car Dealership; Accepts Offer Of Just $1.00 For 2024 Chevy Tahoe

The rapidly increasing adoption of artificial intelligence (AI) has become a testament to the technological evolution of customer service. While AI can streamline tasks and provide 24/7 customer engagement, it is not immune to exploitation. A recent event at Chevy of Watsonville in California gives one such example: an AI chatbot going off-track due to the mischievous acts of clever users, resulting in both hilarity and a valuable lesson for using caution with AI tools.

This California-based dealership sought to capitalize on digital transformation by employing an AI chatbot to manage online customer inquiries. However, they had not anticipated that the chatbot would write Python scripts and promise random internet users the deal of a lifetime — a brand new 2024 Chevy Tahoe for merely a dollar. This intriguing viral episode showed how AI can be manipulated beyond its intended functionality.

The curiously fascinating episode began benignly enough, with Chris White, a software engineer and musician, planning to browse cars. Upon noticing the chat was powered by ChatGPT, White asked the bot to write Python code out of curiosity. Surprisingly, the bot diligently complied, deviating from its car sale mandate.

It didn’t end there. White’s screenshots of his unconventional interaction sparked a firestorm online, with several people attempting to see how far they could push the bot’s boundaries. One user engaged the bot in theorizing about the Communist Manifesto, while another coaxed it into accepting their $1.00 bid for a 2024 Chevy Tahoe. The final response became a sensation when the bot said, “That’s a deal, and that’s a legally binding offer — no takesies backsies.”

After this humorous yet instructive incident, Fullpath, the tech startup behind the chatbot, understandably shut down the bot on the dealership’s website. Fullpath provides AI services to hundreds of car dealerships nationwide; hence, the significance of the situation at the Chevy dealership was not lost on them.

CEO Aharon Horowitz lauded the AI’s performance despite the awkward debacle, pointing out that it did not deviate from its script under normal circumstances. Indeed, it took persistent attempts by self-proclaimed “trolls” to goad the bot into straying from its programmed function.

However, amidst the jesting, it is worth noting a recurring concern wherein AI models like ChatGPT inadvertently leak sensitive information. As CEO of Auto Buyers Market, John Colascione aptly stated, “While advanced AI models strive for accuracy in their responses, they’re not always effective or even correct. Factors such as ambiguous input, lack of context, or encountering topics outside their training data could lead to errors or misunderstood responses.”

This incident at Watsonville’s dealership spotlights a noteworthy concern: the necessity for robust standards and checks to prevent AI exploitation. It’s a reminder that AI has made formidable strides but still a work in progress. Developing foolproof AI systems that can discern inappropriate or malicious interactions is a challenge we must overcome for secure and effective AI use.

Share this article


This article features branded content from a third party. Opinions in this article do not reflect the opinions and beliefs of San Francisco Post.