Beacon Papers No. 21 – The Human Guarantee

Nov 22, 2026 | Papers & Reports, News, Publications

The Authority of AI Must Never Outrun the Capacity to Protect

Nguyen Anh Tuan · Beacon Hill, November 21, 2026

In one week of November 2026, AI’s rewards and AI’s risks moved in opposite directions. Anthropic priced the largest share sale on record, and NVIDIA reported quarterly revenue of $115.8 billion. In the same days, major U.S. banks rewrote their terms so that customers who use AI agents carry the losses when those agents fail. California’s Attorney General sued OpenAI, alleging at least 80 incidents of lost control.

Beacon Papers No. 21 names this the Great Inversion: those with the most power to prevent harm can shift its cost to those with the least. Its answer is one principle, protection before power, and the instruments to make it real:

  • The Human Guarantee. Pay the harmed person first, within days. Settle among the powerful afterward.
  • The Protection Ratio. Every consequential mandate carries verifiable proof of funded protection, valid at the moment of action.
  • The loyal agent. An agent acting for a person must serve that person’s interests, and its provider must answer for that duty.
  • A guarantee without borders. A Human Guarantee Clearing House, beginning in the Founding Cities of AIWS Civilization and reaching Africa.
  • The Capability Dividend. A share of AI’s value is committed to building lasting human capability.

The paper sets out the role of the AIWS Trust Observatory and proposes a 100-day pilot compact for financial agents. It also proposes a new article for the Constitution for Humanity in the Age of Artificial Intelligence: The Protection of Persons from Displaced Risk.

The industrial age learned to insure the deposit. The AI Age must guarantee the person.

Read the full Beacon Papers No. 21 – The Human Guarantee

One person before the house of capital. The authority of AI must never outrun the capacity to protect.
Photo: ANGELA WEISS (AFP)