You're pledging to donate if the project hits its minimum goal and gets approved. If not, your funds will be returned.
$20,000 left to reach the funding goal.
AI agents increasingly hand work to agents from other companies, with no shared way to state what each may do or to prove what happened, so one hijacked or faulty agent can pass harm down a chain no single company oversees. Article 11 AI is building an open protocol, reference relay and test suite that make each agent's boundary explicit and checkable: a published scope, one-use human approval for consequential actions, and a receipt for every action or refusal. This request funds the first three months (November 2026 to January 2027): the threat model and specification v0.1 for public comment, a first public test release, and recruiting outside trial teams. Applications for the protocol's 2027 phases are pending with three AI-safety funders; this is the groundwork those phases build on.
By December 2026: publish a threat model and protocol specification v0.1 (authority labels, provenance, delivery versus agreement, dissent, one-use approvals, receipts) for public comment, in the public domain.
By January 2027: release a first test suite for two of the five boundary failures we track (authorization replay and scope escape), runnable against any agent framework, with the results published.
Line up at least two outside teams to trial the protocol in 2027.
Publish everything under CC0 or a permissive license, with a public log of what the tests catch and miss.
Goal $20,000 over three months: founder time, part-time, $12,000; an outside review of the specification draft, $4,000; model API calls and hosting for the test runs, $2,500; outreach and travel to recruit trial teams, $1,500.
Minimum $5,000: the threat model, specification v0.1 and the first test release at founder pace (API and hosting costs plus part of founder time), without the outside review.
Steve Sonza, founder and CEO of Article 11 AI, Inc., is a retired U.S. Army Major with more than 15 years in Military Intelligence (counterintelligence, signals intelligence, all-source analysis) and an AS in Cybersecurity. Since December 2025 he has run a team of AI agents from several companies under a public-domain constitution (version 2.0, 47 articles), with a human approving every deploy. Article 11 AI runs live at article11.ai: the constitution is readable through a public MCP server and API, two open models hosted on our own GPU answer the public with memory off, and we publish security advisories against our own systems, including our own failures (a time-of-check to time-of-use gap in our authorization instrument, fixed). The company is SBA-certified as a service-disabled veteran-owned small business.
Outside teams may not adopt the protocol, and a one-person company limits engineering speed. If adoption fails, the specification, tests and failure log stay public and reusable by anyone. If engineering is slow, the specification ships first and the tests follow at founder pace.
None. The company has been founder-funded since December 2025. Applications for this protocol's 2027 phases are pending with Foresight Institute, EA Funds' Transformative AI Fund and the Survival and Flourishing Fund; this request covers November 2026 to January 2027, before any of them would start.
This proposal was drafted with AI assistance and reviewed by the founder. Article 11 AI, Inc. is a Wyoming corporation, majority owned and controlled by Steve Sonza, a service-disabled veteran, who serves as its CEO and sole director.
Links: article11.ai/constitution · article11.ai/ai-security · article11.ai/receipt-verifier