Researchers completed one of the most comprehensive studies to date on the risks of using AI models to develop software. In a paper, they demonstrate how a specific type of error could pose a serious threat to programmers that use AI to help write code.
