I have examined real lawsuits of AI agents, and I concluded that without 3 main pillars the most money of most cases will be earned by word jongleurs.
A list of harms AI agents can do
If AI agents work on behalf of human users, then the actions they make are not infinite. As the number of verbs are limited in any language, the number of harmful acts is limited. If you wish you can see a list of potential harms here: https://aiperse.org
If we know what AI agents can do, we can regulate them. Otherwise, do we know what we are talking about?
Clear definitions carved in stone
In many lawsuits people are discussing whether an AI agent is a person or not. Has it got right for free speech? What is an agent at all?
Until we don’t make a list of what NOT AGENTS are, we will never know what agents are. (I know it would be a big alteration to the legal system of the US.)
Without ethical approach you can never judge and regulate this domain
If we push out all ethical considerations from law, there are situations where we don’t have a stable ground to making judgements.
Number one: the user must not use the agent. He is free to make a choice. Yes? No. There will be and are situations where there is no choice. In this case can the user be responsible for what he does? Is it a forced action?
Number two: if the user is targeted by marketing tools to do something that he regrets later and he can prove that he was tempted, who is responsible? Who is more responsible: the tempter or the tempted?
I have written this all alone without using any AI tool or agent. But I must cote Seyfarth about my responsibility. The text, written by me, Piroska Rakoczi “should not be construed or relied on as legal advice or to create a lawyer-client relationship. Readers should not act upon this information without seeking advice from their professional advisers.” (source: https://www.seyfarth.com/news-insights/mobley-v-workday-court-holds-ai-service-providers-could-be-directly-liable-for-employment-discrimination-under-agent-theory.html)