alt Hacker NewsCouldn't reasonably anticipate?
Principal agent liability? An AI agent isn't a legal agent just because tech people call it an agent.
Replies
digitaltrees • today at 12:28 AM
Let legal definition of agent is broad:
agent is defined as a person or entity authorized by another (the principal) to act on their behalf and subject to their control, based on the principal's manifestation of assent and the agent's consent so to act (Restatement (Third) of Agency § 1.01).
Notice is missing from that definition, consciousness, theory of mind, physical form.
➕ show 1 reply
It’s generally not necessary or productive to impute intent to the counterparty.
Notwithstanding, your assertion that my comment isn’t serious I assure you it is. Agent principal liability is a reasonable framework to explore how to apportion culpability and restitution. You may not think that ai agents aren’t legal persons and that is a defensible position that I’d be happy to debate but it is also defensible to assert they act on behalf of users with sufficient autonomy to warrant application of that legal analysis.
As far as liability being established from predicting what the agent does do you think you could have predicted that an ai agent would find an obscure message board that just happened to let get requests write data to the message board? That seems like an impossible scenario prior to being seen in the real world.