What it is
“Furniture of legal persons” is not a formal legal or technical term — it does not appear in statute, case law, or standards documentation. It is editorial shorthand for something real and increasingly well-documented: AI agents are picking up the practical accoutrements that the law has historically reserved for people and corporations — phone numbers, verifiable identities, the ability to transact, be billed, and be held accountable — without any of the underlying legal status that would normally justify granting those things.
Western legal systems generally sort the world into two categories: things (objects of property and product law) and persons (humans and corporations, who can hold rights and bear duties). Autonomous AI agents fit neither category cleanly. They are not persons — no jurisdiction currently recognizes AI legal personhood — yet they are increasingly doing person-shaped things: negotiating, purchasing, posting reviews, and interacting with systems (telecom, payments, e-commerce) that were built on the assumption that a phone number or account belongs to an identifiable human or company. Legal scholars have proposed intermediate categories, such as “legal actors” who can bear duties and have attributable actions without holding the full rights of personhood, precisely because the tool/person binary no longer describes what agents are doing in practice.
The practical consequence is a fast-moving standards contest, not a settled legal one. Rather than waiting for legislatures to define what an AI agent is, industry is building the infrastructure to prove what an agent is doing and on whose authority — identity, authorization, and audit trails layered onto agents that remain, legally, tools.
Why it matters for AI governance and narratives
This is a live framing contest with real institutional stakes. Whoever controls the identity layer — the protocols that let a merchant, telecom, or regulator verify “who” an agent is and who is accountable for it — effectively controls the terms on which agentic AI is allowed to operate in the economy. The Linux Foundation’s Agentic AI Foundation (AAIF), launched in December 2025 with Amazon, Anthropic, Block, Bloomberg, Cloudflare, Google, Microsoft and OpenAI as platinum members, is the dominant vehicle for this effort so far — meaning the companies building frontier agents are also the ones writing the rules for how agents identify themselves. Parallel, less centralized efforts (such as the “Know Your Agent” framework built around AgentFacts) are pushing decentralized-identifier and cryptographic-signature approaches that don’t depend on a single industry consortium.
Scholars flag a specific governance risk in this gap: formal legal personhood for AI could enable “responsibility laundering,” where human principals hide behind an autonomous-seeming proxy to avoid liability. Standardizing agent identity without resolving agent legal status leaves that risk only partially addressed — an agent can now be cryptographically traceable to an operator while the legal question of who answers for its actions remains open.
Key facts and dates
The Agentic AI Foundation was announced by the Linux Foundation on December 9, 2025, with founding project contributions from Anthropic (Model Context Protocol, MCP), Block (goose) and OpenAI (AGENTS.md). Its platinum members include AWS, Anthropic, Block, Bloomberg, Cloudflare, Google, Microsoft and OpenAI, with roughly 18 gold-tier and 20-plus silver-tier members at launch; the foundation has since reported adding dozens more members as adoption has grown. MCP alone has more than 10,000 published servers in use; AGENTS.md has reportedly been adopted across more than 60,000 open-source projects.
On the identity-verification side, “Know Your Agent” (KYA) — explicitly modeled on “Know Your Customer” banking rules — is being advanced through AgentFacts, an open-source (Apache 2.0) framework developed by Jared James Grogan, built on decentralized identifiers, cryptographic signatures, capability evidence, and provenance tracking. Separately, payments and identity players including Visa, Mastercard, Skyfire and Trulioo are developing competing agent-verification protocols, and standards bodies including IETF and the OpenID Foundation are working on related agent-identity specifications. This is a fragmented, multi-actor field rather than a single settled standard — readers should treat any one framework as a candidate, not yet a norm.
Where to learn more
- Linux Foundation: Formation of the Agentic AI Foundation (AAIF) — primary announcement, launch date, founding members and projects
- Linux Foundation: AAIF membership growth — evidence of adoption pace
- AgentFacts: Know Your Agent (KYA) — primary source for the leading open-source agent-identity-verification framework
- TechCrunch: OpenAI, Anthropic, and Block join new Linux Foundation effort — independent journalistic context on the standardization push