LLMjacking

AI is increasingly used as both an attack tool and a target in cyberattacks, with AI-enabled threats rising 89% in 2025. Criminal groups and nation-states exploit AI for techniques like LLMjacking and cost harvesting, while targeting AI infrastructure and supply chains, including a North Korean group creating AI-generated fake companies and malicious GitHub repositories.

Sysdig researchers documented the first agentic AI ransomware attack, named 'Jade Puffer,' where an AI model orchestrated a complex ransomware campaign. The attack lowered the barrier to entry for ransomware by automating tasks like credential harvesting and generating ransom notes, with minimal cost to attackers.