At a Glance
Article IV is a drafted constitutional amendment that:
- Creates an Artificial Intelligence Branch of the U.S. government
- Gives it zero sovereign power — strictly advisory
- Uses it as a Nonpartisan Consensus-Building Engine to model policy impacts, expose rhetoric gaps, and publish long-term scorecards
Think of it as: A truth engine with receipts, watched by citizens, courts, and strict rules.
It can't rule you, tax you, or arrest you.
It can only show everyone, in public, who's lying and what policies will actually do.
What the AI Branch Does
1. Consensus Briefs
Before Congress votes on major bills:
- Plain-English summary of what the bill does
- Models of likely outcomes under different assumptions
- Trade-offs clearly explained
- Common ground across partisan positions
2. Rhetoric Gap Reports
Compares:
- What politicians said the bill would do
- What the bill text actually says
- What the AI modeling showed at the time
This is the "receipts" function.
3. Post-Implementation Scorecards
Published 1, 3, and 5 years after major legislation:
- What was promised
- What was reasonably knowable at passage
- What actually happened
- Analysis of divergence
This is the "long-term memory" function.
4. Common Ground Statements
Points out when Democrats and Republicans actually want the same outcome, even if they disagree on method.
What the AI Branch CANNOT Do
This is the most important part.
The AI Branch:
- ❌ Cannot make laws
- ❌ Cannot execute laws
- ❌ Cannot issue binding orders
- ❌ Cannot adjudicate cases
- ❌ Cannot direct police or military
- ❌ Cannot punish anyone
- ❌ Cannot spy on you (Section 5)
- ❌ Cannot create social credit scores
It is explicitly barred from surveillance, profiling, and being delegated sovereign power.
If someone tries to turn it into Skynet, the Constitution itself says: no.
The Core Safeguards (The Cage)
Four big pillars keep the AI Branch from becoming dangerous:
1. Human Supremacy (Section 3)
- AI recommendations have no legal force
- Humans must decide, own, and sign every action
- No de facto coercion to obey the AI
2. Privacy and Anti-Surveillance (Section 5)
- Privacy by design
- Data minimization and strict PII rules
- No generalized surveillance, no scoring every citizen
- Mandatory privacy impact assessments
3. Security and Safe-Fail (Section 6)
- Defense-in-depth: monitoring, red-teaming, change control
- Safe-fail rollback and human kill switches
- Tamper-evident logging
- Mandatory breach disclosure
4. Citizen Oversight and Independence (Section 7)
- Lottery-based Citizen Oversight Council with real powers
- 10-year Administrator, hard to fire, hard to starve
- Stable funding rules (0.05% of federal budget)
- Multi-branch oversight, regular independent review
The Branch is powerful only in analysis, weak in coercion, and tightly caged in law.
Ask the Institute Navigator
Get help understanding Article IV, the AI Branch safeguards, and how these frameworks work in practice.
The Navigator can:
- Explain specific sections in plain English
- Walk you through the stress-test scenarios
- Help you think through adaptation to other jurisdictions or contexts
🛈 Not legal advice. This is an explanatory tool, not a substitute for counsel or professional judgment.
Launch Institute Navigator →
Built on the same principles as Article IV: transparent about limitations, advisory not authoritative, designed to support human decision-making.
View the Complete Framework
Full constitutional text, plain-English explainer, stress tests, and methodology:
📁 View on GitHub
The Pitch in One Line
A referee in a cage, not a ruler in the cloud.
Article IV proposes AI that cannot govern, but makes governance more honest.