Generative Artificial Intelligence and biased responses: What we should pay attention to when combining journalistic work with ChatGPT

The article approaches, with a critical eye, the use of Generative Artificial Intelligence as a tool for journalism, supporting the hypothesis that the use of technology should be done with caution due to machine biases. The objective is to identify ChatGPT's ability to produce journalistic content and verify the existence of three types of biases. The methodology employs Content Analysis to examine one text, and three images created by GPT-4 from a real news story. The results indicate a high incidence of false answers (72%) and hallucinations (27%) in the texts, in addition to false bias (33%) and prejudice bias (100%) in the images. It is concluded that the use of generative AI in journalism must be judicious, with the need for human supervision to avoid the perpetuation of misinformation and prejudice, in addition to a clear need for ethical regulation for the use of the technology.