The Authority Gap

AI agents operate at machine speed. Halt authority remains human.

0
Agents with automated halt authority
56,308
Agents with human halt authority
39,568
Agents with no classified authority

Of 95,876 AI agents across 543 regulated financial institutions, 56,308 have a named halt authority. Every named authority is human: an operator, a manager, a committee, or a board. Zero agents have an automated halt authority that engages without human action. Halt is hierarchically gated in a system whose agents transact at machine speed.

Figure 1 · Halt Authority Distribution Across 95,876 Agents
0 10k 20k 30k 40k AGENTS 0 AUTOMATED 25,547 26.6% OPERATOR 21,251 22.2% MANAGER 9,061 9.5% COMMITTEE 449 BOARD 39,568 41.3% UNKNOWN HALT AUTHORITY Where halt authority is named, the named authority is always a human. MAR®500.com

Each bar groups agents by who holds authority to halt them. The leftmost bar (automated, dashed at zero) marks the absence: no agent in the substrate has a halt authority that engages without human action. The next four bars show the human chain in descending order of frequency: operator, manager, committee, board. The rightmost bar collects the 41 percent of agents whose halt authority has not been classified at all. The Coase Inversion is visible here in operational form: every named authority needs a human to act, and human action lags machine-speed transactions by orders of magnitude.

Source · Meridian substrate v13.1.0, May 2026 · halt_authority field · All evidence tiers Methodology · Cascade Without Containment (working paper)
543 institutions · 12 sectors · 95,876 agents · 636,854 governance edges · substrate v13.1.0
Methodology grounded in Cascade Without Containment, working paper. Substrate methodology in The Stationary Sea (Part 1) on Zenodo.