ChatGPT generates mostly insecure code, putting user data at risk

OpenAI's Chatbot, ChatGPT, has been found to produce mostly insecure code that fails to meet minimal security standards, according to computer scientists at Universite du Quebec. ChatGPT can identify security problems but does not alert users to its inadequacies, leaving them at risk. Further prompting allows the AI model to produce more secure suggestions; however, the Chatbot does not assume an adversarial model of code execution, which leads to vulnerable code creation. The authors suggest improving ChatGPT's ability to provide useful guidance to make generated code secure.