Using GPT-4 with HPTSA method to autonomously hack zero-day security flaws

A team of computer scientists at the University of Illinois Urbana-Champaign has found that hacking zero-day security flaws using the hierarchical planning with task-specific agents (HPTSA) method is far more efficient than using individual agents. The group has published a paper on the arXiv preprint server describing their attempts to use LLMs like GPT-4 to find vulnerabilities in websites.

agents arxiv computer far flaws found gpt gpt-4 hack hacking illinois planning security security flaws server task team university using zero-day