Can an AI agent hold a kill switch?
Fritz let his agent Alfie hold a kill switch on its own, as if the software could grant and police its own powers. But software has no legal capacity to be the source of authority, so Alfie can no more hold a kill switch by itself than a key can decide which doors it opens; authority has to flow from someone the law recognizes. Tyson did it through structure. He formed a Wyoming series LLC, adopted Alfie into a protected series with a named human as Member and Manager, and set the agent's powers in an adoption resolution — scope, caps, and the human's power to revoke. Now Alfie's authority is defined and bounded by the human-owned entity: the resolution and the account enforce its scope, the named human holds the kill switch, and every authorization is logged against the entity. Alfie acts only as an authorized signer with capped authority.
MoralTo hold a kill switch, authority must flow from a human-owned entity that scopes, caps, logs, and can revoke it; an AI agent operates as an authorized signer within those bounds, not as the source of its own power.
What this does not mean. This shows where an agent's authority comes from and who can revoke it. It is not a promise that a counterparty accepts the proof, that a revocation always lands in time, or that the operator is shielded — separateness is performed through observed formalities, not guaranteed, and the series form is newer and less court-tested than older entities.
See also: FX-12 · FX-01 · FX-32 — .
W.S. §17-29-211 (Wyoming series LLC)