Researchers recently demonstrated that their artificial intelligence model, GPT-4, can autonomously hack one-day vulnerabilities with an impressive success rate. In a subsequent study, they showed that by employing a team of specialized agents through a unique HPTSA technique, they were able to hack zero-day vulnerabilities even more efficiently. This approach is reminiscent of how real-world software development teams operate and proved to be significantly more effective than using a single AI agent. However, there are concerns about potential misuse of this technology for malicious purposes. Despite its capabilities, GPT-4 in chatbot mode reassured users that it operates within ethical and legal boundaries.
https://newatlas.com/technology/gpt4-autonomously-hack-zero-day-security-flaws/