Philosophical question to ponder about AI writing programs

[TheRealNorth]

If the AI commits a crime or does something that causes damage (for instance, commits libel), who is responsible, criminally or financially?

In a civil trial, I could see the person that executed the script as being liable for financial damages.

[Ms. Toad]

1. No different than using any other tool.

If you use a knife to kill someone, you - as the person wielding the knife are liable for the killing.

If you use computer software to commit bank fraud, you - as the person devising the scheme which uses the software, are liable for the fraud.

AI is a tool. If you use it to cause harm you will be responsible for the harm (assuming it falls within the elements of a civil claim or criminal charge).

As to copyright infringement, it is relatively untested law. But it is likely that both the entity that provided the training database (reproducing the work in a traininig form) and the person who used the software will be liable (and, in doing so created a derivative work from the original works). Assuming that, once the law is settled, using work for training without licensing it constitutes infringement. (It should, but it's relatively new law.)