Problem of Using Generative Artificial Intelligence in the Development of Software Products

In this contribution, we attempt to analyze the primary problems associated with the use of generative artificial intelligence technologies in software development. We consider the limited interpretability of results, which is a major challenge for developers when debugging and ensuring the program security. Reliability problems associated with variability of input data, issues of quality of training data and their impact on generation results are also topical issues of this research. A large pool of problems is associated with the ethical aspect: bias and discrimination, issues of responsibility for the results created with the participation of the GenAI. In the context of these issues, we come to the need to develop standards and methodologies that can ensure the interpretability, reliability and ethical use of GenAI in further software development. This work is intended to be a starting point for deeper discussion and research in this area. The conducted analysis is based on research results and analytical reviews of large consulting and IT companies.