Athira Sethu
Kochi, 27 December 2024
ChatGPT search can be fooled into giving misleading results, reports said. According to a recent release by The Guardian, chatGPT Search, an AI-powered search engine out this month, can also be easily manipulated to pass on misleading information.
ChatGPT Search is programmed to make browsing faster as it summarizes information, be it product reviews, etc., to enable quick decisions. However, The Guardian found that by adding this hidden text to websites, it was possible to hoodwink ChatGPT into ignoring negative reviews while creating summaries that were solely positive. In some cases, this method also caused the AI to generate harmful code.
This is a known problem for AI models like ChatGPT, but it is the first time it has been seen in a live AI search tool. While Google, a leader in search, has more experience in handling such problems, this new discovery highlights potential risks with AI-powered search engines.
