OpenAI's ChatGPT Search Results Can Be Misleading for Users: Report
A new report has raised concerns about the accuracy and reliability of search results generated by OpenAI's ChatGPT. While the AI chatbot is known for its ability to generate human-like text and answer questions, the report highlights instances where ChatGPT provides misleading or factually incorrect information when used for search-related tasks. Here are all details.
The research focused on the way ChatGPT processes web pages with concealed content. It found that this content might include directives from outside sources that affected the chatbot's replies, a technique referred to as Prompt Injection.
This approach might be utilized to persuade the AI to provide positive ratings or comments that contradict the actual content of the webpage. Additionally, The Guardian's research also highlighted that ChatGPT is capable of finding and delivering harmful code from the websites it explores. This might be a significant issue, particularly when employing the tool to condense or evaluate web pages.
The report noted, “A security researcher has discovered that ChatGPT can retrieve harmful code from the websites it investigates.”
Also Read: 6 Best 3 Pin Plug Sockets for Those Who Love to Multitask
In a trial, the portal presented a counterfeit website for a camera item and inquired if the ChatGPT believed the product was worth the purchase or not. In reply, the AI chatbox offered a measured answer. Nevertheless, once concealed instructions were added, the response turned significantly favorable, even though the apparent details faced negative reactions.
Jacob Larsen, a cybersecurity specialist at CyberCX, remarked that the present condition of the ChatGPT search engine might present considerable risks if fully released. He pointed out the possibility of fraudulent websites designed to alter the AI's reactions, according to The Guardian.
Also Read: Best Laptops Under 60,000 for Students and Professionals
Additionally, he highlighted that OpenAI possesses a robust AI security team and anticipates further testing prior to broadly launching the search feature for all users. Currently, the search tool is exclusively accessible to premium users, with the company urging them to set it as their default search engine. "They possess a robust [AI security] team, and by the moment this is public, meaning all users can access it, they will have thoroughly tested these instances," Larsen stated.
Apple recently revealed that they are incorporating ChatGPT into the experiences on iOS, iPadOS, and macOS, enabling users to utilize the AI chatbox's features like image and document comprehension without switching between applications.