Malicious Manipulation of Large Language Models in Automated Exploit Development

The article warns that generative AI and large language models are being repurposed by adversaries to automate vulnerability discovery and exploit development through techniques like data poisoning, adversarial prompting, and model inversion; it describes the emergence of autonomous exploit pipelines that can scan for flaws, generate payloads, and weaponize findings at scale, and recommends defensive responses including targeted defensive models, continuous auditing, and monitoring for exploit-like outputs.