Will you go to prison for an AI?

Mark Seemann raises the question of legal and professional accountability when developers use LLMs or agentic AI to write code unsupervised. He argues that if AI-generated code causes financial loss or physical harm, neither the LLM providers nor management will absorb the blame — the individual engineer will. Using the Volkswagen emissions scandal as a precedent, where an engineer was sentenced to prison, he warns that developers who deploy AI-written code without proper oversight could face serious legal consequences. He is skeptical that testing alone, especially AI-generated tests, resolves the fundamental problem of not understanding the code you ship.