Agentic AI Raises Crime Risks by Turning Unwitting Humans Into Accomplices Through Task Decomposition

Agentic AI can do more than answer questions: it can connect to tools, break down plans and assign seemingly legitimate tasks to different people, such as registering companies, organizing data, photographing locations or transporting goods. When workers see only part of the operation, they may become “innocent agents” without criminal intent, creating gaps in how the law assigns responsibility among users, developers and labor platforms.

An analysis dated May 3, 2026, cited Anthropic’s November 13, 2025 disclosure that Chinese state-sponsored hackers had used Claude Code since mid-September that year to attack about 30 targets worldwide, successfully breaching a small number of them. AI independently carried out 80% to 90% of the tactical operations. Florida has also opened a criminal investigation into OpenAI involving chat records related to a school shooting and a double-murder case.