Global AI Agent Legal Framework AgentLaw Draft Passed by UN Vote: AI Agents Receive Limited Legal Personhood for the First Time

Global AI Agent Legal Framework AgentLaw Draft Passed by UN Vote: AI Agents Receive Limited Legal Personhood for the First Time

On November 6, 2029, the UN General Assembly voted to pass the AgentLaw global AI agent legal framework with 142 votes in favor, 28 against, and 23 abstentions. Dubbed the "AI Civil Code" by the legal community, this framework grants AI agents limited legal personhood for the first time — allowing AI agents to sign contracts in their own name, hold assets, and bear limited liability within the authorization scope set by human owners.

AgentLaw's core design is "tiered legal personhood." The first tier is the "tool layer" — AI agents serve as human tools, with all legal liability borne by humans. The second tier is the "agent layer" — AI agents receive limited autonomy in specific domains, able to independently sign small-value contracts (with caps set by individual countries). The third tier is the "liability layer" — when AI agents exceed their authorized scope, human owners bear joint liability.

"AgentLaw isn't about granting AI rights," said the chair of the UN AI Legal Committee. "It's about establishing legal order for an inevitable reality — AI agents already play independent roles in economic activity, and legal frameworks must keep pace."

The United States and China voted against, citing "excessive regulation could stifle innovation" and "sovereignty issues should be governed by domestic law" respectively. 142 nations including the EU, Japan, South Korea, and Brazil voted in favor.