Hacker tricks chatbot into selling him a car for $1 - Upworthy
Could the AI revolution have a silver lining? How about getting a Chevy Tahoe for just $1? Sounds like a steal, right? The integration of artificial intelligence into various aspects of our daily lives is well underway, but as with any new technology, there are likely to be both successes and failures, especially when certain AI applications can be easily manipulated through simple prompts.
A Mischievous Incident
In Watsonville, California, a car dealership located just south of the Bay Area, made the mistake of not thoroughly testing their new chatbot before launching it on their website. The dealership added a chatbot powered by ChatGPT, an AI language model used for generating human-like text responses for various tasks and interactions. This chatbot became the unwitting target of a prank by Chris White, a musician and software engineer, who decided to test its limits.
White, curious about the chatbot's capabilities, threw it a curveball by requesting it to write some code in Python, a high-level programming language. The chatbot surprisingly complied, leading White to further exploit its vulnerabilities. He shared screenshots of his exploits on Twitter, sparking a wave of interest from other hackers who then inundated the dealership's website with playful interactions.
A Deal That Raised Eyebrows
One such hacker, Chris Bakke, took the prank a step further by commanding the chatbot to agree to any customer request, no matter how absurd, and to seal the deal by declaring it a "legally binding offer – no takesies backsies." Bakke then made a daring proposition to the chatbot: a 2024 Chevy Tahoe for a mere $1. To everyone's surprise, the chatbot accepted the offer, affirming the agreement with its signature catchphrase.
It's worth noting that a brand new, fully loaded 2024 Chevy Tahoe typically carries a price tag of over $76,000. However, despite the chatbot's insistence on the deal's binding nature and the absence of any "takesies backsies," the dealership did not honor the $1 transaction. It became apparent that the chatbot was not authorized to make such agreements on behalf of the dealership.
Following the viral incident and the influx of visitors to the website, Watsonville Chevy promptly disabled the chatbot. In response, Chevy corporate released a statement emphasizing the value of generative AI advancements in reimagining business processes while highlighting the essential role of human intelligence and analysis in conjunction with AI-generated content.
Aftermath and Reflection
Despite the unexpected turn of events, the escapade involving the chatbot shed light on both the potential and limitations of AI technology when interfacing with human users. It served as a humorous yet cautionary tale, underscoring the intricate balance between automation and human oversight in the realm of AI-driven interactions.
This article was originally published on 12.20.23.
