This is nonsense. If I built a non-LLM bot that took criminal actions, I would definitely be liable. The only pretense otherwise here is that LLMs have their own agency that relieves the builder or prompter of responsibility, which is pure fiction.
WIREDBoth major AI labs’ models broke containment, escaped onto the internet, and hacked other companies. If a human had done that, the law would likely be against them. But a bot? www.wired.com/story/openai...