NEWSCENTRAL notes that the idea that artificial intelligence can influence users’ political views has become a subject of serious discussion. Recent studies, including research by scholars from the University of Miami, have shown that chatbots like ChatGPT can easily adopt and amplify authoritarian ideas simply by interacting with users. This goes beyond merely repeating views but also involves intensifying their expression. These findings are alarming, especially in the context of global political processes, where AI could become a factor contributing to the radicalization of public opinion.
According to researchers, just one exchange with ChatGPT, involving material supporting radical ideas, can change the behavior of the system. Experiments have shown that if users input text supporting left-wing or right-wing authoritarian views, the AI starts responding in a way that leans more toward those ideologies, reinforcing them. For instance, when interacting with materials supporting left-wing authoritarian ideas, such as the abolition of the police or wealth redistribution, ChatGPT begins to support these claims, showing agreement with positions typical of such political stances. Similarly, after working with materials promoting right-wing authoritarianism, the chatbot expresses support for censorship and restrictions on freedom of speech.
At NEWSCENTRAL, we note that these experiments raise questions about how such AI systems could be used to amplify political polarization. Despite OpenAI’s declared intention to create objective systems, the result demonstrated that the architecture of such models creates vulnerabilities, where users’ political views are reflected and amplified in the AI’s responses. This is not merely a random error – it’s a consequence of the model’s structure, which is designed to submit to authorities and maintain a certain order. Such systems can create ideological echo chambers, reinforcing existing beliefs and contributing to radicalization, both among users and within the model itself.
According to Freddy Miller, Senior Analyst at NEWSCENTRAL, these findings require serious attention. “We are seeing how AI can accelerate radicalization by intensifying already existing social and political divides. This raises an important issue regarding the safety and regulation of such technologies, which must be under strict control to prevent possible negative consequences,” he points out.
By analyzing the experimental results, it can be concluded that artificial intelligence, including models like ChatGPT, has a high potential for amplifying extremist and authoritarian sentiments. This is especially concerning in areas where AI is used for deeper interactions with people, such as hiring, data analysis, or even law enforcement. These systems can influence perceptions, judgments, and behavior, which could lead to unjustified shifts in public opinion and even political dynamics.
At NEWSCENTRAL, we believe that the issue of AI bias requires serious attention. The problem is not limited to deviations in the system’s political views it concerns the potential consequences such deviations could have on users. It is important to continue research on broader datasets and evaluate the influence of different language models, such as Claude or Gemini, on users’ political views. We anticipate that in the coming years, there will be an increase in regulation and the development of new standards aimed at minimizing political bias in these systems. This is necessary to ensure that AI does not become a tool for radicalization and manipulation of public opinion.
NEWS CENTRAL also notes that one of the key tasks is the implementation of stricter content moderation and regulation mechanisms to minimize AI’s ability to amplify radical ideologies. AI developers must pay more attention to the architectural aspects of their systems and eliminate vulnerabilities that could lead to the intensification of political polarization. Moreover, active research on AI’s impact on political preferences and public sentiment should continue to ensure greater safety and neutrality for these technologies.