Contrary to the law of headlines the answer is yes. > Overall, we find that participants who had access to an AI assistant based on OpenAI's codex-davinci-002

Contrary to the law of headlines the answer is yes. > Overall, we find that participants who had access to an AI assistant based on OpenAI's codex-davinci-002 model wrote significantly less secure code than those without access.