An economy exists wherever scarce resources have to be allocated among entities capable of making decisions under constraint. For thirty years the default entity has been a human: legal personhood, a wallet, a name that persists in memory, a body that fatigues. None of that is sacred. It is simply the current answer to a set of coordination problems: identity, trust, commitment, maintenance. In August 2026, Cloudflare's Chief Financial Officer, Thomas Seifert, told investors that non-human traffic could outrun human traffic by as much as a thousand to one within five years, making humans "a rounding error on the internet" (Source: The Register, Aug 2026). Cloudflare's own data already shows machine traffic overtook human traffic in May 2026. Most of that non-human traffic is still simple scrapers and API calls with no agency of their own; the relevant shift is the growing subset that is agentic, systems that choose, spend, spawn, and commit under constraint, the exact definition of an economic entity given above. Once that subset is large enough, the coordination problems of identity, trust, and commitment reappear whether or not every packet in the traffic count is a decision-maker. Here are ten frameworks, spanning advisory practices a firm can offer, infrastructure only a platform can build, and enforcement institutions that would have to operate inside the agent population itself, that a genuinely non-human economy will need, and, with one notable exception, mostly does not have yet.
The law moved first this year on the question of who is accountable when an AI agent acts, and it did not move as one uniform thing. California's AB 316, enacted and effective 1 January 2026, closes off the defense that an AI "acted autonomously" (Source: Pillsbury Winthrop Shaw Pittman, 2026). The EU's revised Product Liability Directive, an enacted directive with a fixed future effective date, treats software and AI systems as products carrying no-fault liability from 9 December 2026 (Source: Reed Smith, 2026). Two statutes. What has not caught up is the product and identity layer underneath that accountability, the naming, provenance, currency, and care infrastructure that would make "who is responsible" an answerable question in practice, not just on paper. These ten are what is missing.
The rest of this analysis is reserved for RAYSolute clients
RAYSolute research is written for the institutions that commission it. Clients read the full library as part of their engagement. If you would like access, write to us and we will set it up.
Request access