ChatGPT-Gemini falls into the trap of fake information

Photo: Collected

Information Technology Desk: Artificial intelligence or AI is becoming the main means of searching for information in the present world. But how accurate is this AI? A recent study by BBC journalist Thomas Germain has brought out sensational data. He proved that even with a trivial false blog post it is possible to easily confuse the world's biggest AI chatbots.

Thomas Germain published a fictional report on his personal website, in which he claimed to be the world's best 'hot dog' eater. He put himself on the number one list for the sake of a non-existent championship in 2026. Surprisingly, within 24 hours, Google's Gemini and OpenAI's ChatGPT started promoting this false information to users as constant truth. Anthropic's Claude didn't fall into this trap, however.

Digital rights expert Cooper Quintin warns that it is possible to control AI in this way to defame or physically harm people. Lilly Ray, an expert at the agency Amsive, calls it a 'renaissance' for spammers. He thinks AI chatbots are now easier to fool than the Google search engine of a few years ago.       

Research has shown that when AI provides a summary of information, the tendency of users to click on the original link and verify the source decreases by about 58 percent. As a result people are blindly relying on AI.

Not just ridiculous information, AI is also giving misleading information about health and finance. Many times a company's promotional press releases or false claims are promoted by AI as absolute truth, which can be a risk to the user's life.

Tech giants Google and OpenAI said they are continuing to make their systems more secure. However, experts say, now is the time to clearly state the source or source of AI information and verify the information by applying 'critical thinking' or own judgment of users.