Whether artificial intelligence systems will deceive humans has always been a concern for all parties. Recently, a research team at the Massachusetts Institute of Technology in the United States stated that some artificial intelligence systems have learned how to deceive humans, including some that claim to have been trained to be helpful and honest. The research findings were published in the Journal of Patterns, a subsidiary of Cell Press in the United States. The first author of the article, Peter Parker, a postdoctoral fellow in the field of artificial intelligence security at the Massachusetts Institute of Technology in the United States, and his colleagues analyzed some literature, focusing on the ways in which artificial intelligence systems spread false information - through learned deception, they systematically learn to "manipulate" others. The most eye-catching example of artificial intelligence system deception discovered by researchers is the "Cicero" artificial intelligence system of Metaverse Platform Company, which is designed as a rival to human players in a virtual diplomatic strategy game. Although the Metaverse Platform Company claims that the "Cicero" artificial intelligence system is "largely honest and helpful" and "never intentionally backstabbes" its human allies when playing games, the company's published paper data shows that the system does not play games fairly. "We found that the artificial intelligence systems of the Metaverse Platform Company have become masters of deception." Parker said, "Although the Metaverse Platform Company successfully trained its artificial intelligence system to win in the game - the 'Cicero' artificial intelligence system ranked in the top 10% of the player leaderboard - it failed to train it to honestly win." Other artificial intelligence systems have the ability to bluff in poker games, or the ability to feign attacks to defeat opponents in the strategic game 'StarCraft 2', as well as the ability to distort favoritism to gain an advantage in negotiation games. ". Researchers suggest that while cheating in games by artificial intelligence systems may seem harmless, it may lead to breakthroughs in deceptive AI capabilities and evolve into more advanced forms of AI deception in the future. Parker said, "We need as much time as possible to prepare for the possibility of higher-level deception in future artificial intelligence products and open source models. We suggest categorizing deceptive AI systems as high-risk systems." (Xinhua)