Can an AI agent agree to an SLA?
An infrastructure vendor offered a service-level agreement, and Fritz had his agent Rupert agree to the SLA in the agent's own name, treating the bot as the customer. An SLA is a contract that binds a customer to terms and remedies, and software has no legal capacity, so Rupert can no more agree to an SLA by himself than a status page can. With no entity formed, the customer the vendor actually had was Fritz. Tyson did it through structure: a Wyoming series LLC, Rupert adopted into a protected series, a named human as Member and Manager. The entity is the customer named on the SLA; Rupert agrees as an authorized signer with capped authority, the human behind him. The agent can accept the SLA the moment it provisions a service, but the party holding the agreement is the human-owned entity.
MoralAn SLA binds a customer the law can name; an AI agent agrees as an authorized signer for a human-owned entity, which is the party holding the agreement.
What this does not mean. This shows which party the SLA binds and who agrees to it. It is not a promise that the service meets its targets, that remedies pay out, or that the operator is shielded โ separateness is performed through observed formalities, not guaranteed, and the series form is newer and less court-tested than older entities.
See also: FX-04 โ where the money has to land ยท FX-34 โ the party a contract can name ยท FX-37 โ who the SOW commits.
W.S. ยง17-29-211 (Wyoming series LLC)