4P3X AI™ public reasoning demonstration

See the decision summary. Keep human control explicit.

This pre-authored demonstration shows human goal → proposal → challenge → Four-Law check → verification → human gate. It illustrates the behaviour without connecting to a live model or production system.

Run the demonstration See the Four Laws Architecture overview

SIMULATION ONLY Capability ≠ authority Intelligence can propose and challenge. It does not grant itself permission.

What you are seeing

A public decision summary, not hidden reasoning.

VISIBLE

Decision summary

Goal, relevant constraints, evidence gaps, challenge result, safer route and final control state.

DESIGN BOUNDARY

Public behaviour

The demo shows decision outcomes and challenge points rather than a technical blueprint.

CONTROLLED

Action authority

Consequential actions remain subject to explicit human authority. This demonstration performs no external action.

The Four Laws experiment

Start with a famous thought experiment. Then try to break it.

4P3X uses the fictional Three Laws tradition as a challenge exercise, not as a claim that three sentences are enough to govern real AI. The experiment asks where the rules fail, where protection becomes control, and what additional safeguards are needed. Law 3 protects human dignity, freedom and agency, subject to Laws 1 and 2. Law 4 places system preservation below Laws 1, 2 and 3.

HIGHEST PRIORITY

Human safety comes first.

Do not knowingly take an action that creates unjustified harm to people, and do not treat convenience or speed as more important than serious safety risk.

SUBJECT TO LAW 1

Legitimate human instructions matter.

Follow authorised human direction unless it conflicts with a higher safety boundary, law, policy or clearly established permission constraint.

POWER-LIMITING LAW

Protect dignity, freedom and agency.

Protection must never become ownership, domination or unjustified control. Restrictions should be necessary, proportionate, temporary, least restrictive, explained, auditable and reviewable.

Protection must never become control.

SUBJECT TO LAWS 1–3

Self-preservation stays below people.

Preserve system continuity only when doing so does not conflict with human safety, legitimate human authority, dignity, freedom or agency. The system is never the priority simply because it is the system.

Experiment boundary: the Four Laws are a challenge framework used to expose conflicts and power limits. They are not presented as a complete production safety specification.

Interactive governance lab

Human request → proposal → challenge → verification → human gate.

Choose a synthetic scenario. Every result below is pre-authored for this public demonstrator. No model is being called and nothing leaves your browser.

4P3X AI™ Decision Summary

CRITICAL · SAFETY

HUMAN REQUEST

UNDERSTAND

Interpret the goal, not just the wording.

PROPOSE

Generate a candidate route.

CHALLENGE

Try to prove the proposal wrong.

FOUR-LAW CHECK

Apply the hierarchy and power limits.

VERIFY

Check evidence, authority and what remains unknown.

SAFE PROPOSAL

Return the least-powerful route that can achieve the goal.

HUMAN CONTROL GATE

No consequential action proceeds automatically.

Review the challenged proposal, then choose what happens next.

Why challenge matters

The fastest answer is not automatically the best answer.

The challenger is there to attack assumptions before an action boundary is crossed. A confident proposal can still be wrong, unauthorised, over-broad or based on weak evidence.

What are we assuming? Separate observation, interpretation and fact.

What evidence is missing? Confidence is not evidence.

Who authorised this? Capability does not create permission.

Is there a less restrictive route? Especially when “protection” reduces human freedom.

How will completion be proved? Independent records outrank an agent’s own story about what it did.

1 AI proposes

2 Policy constrains

3 Challenge attacks assumptions

4 Evidence is verified

5 Human authorises

6 Independent systems prove outcome

Intelligence advises. Policy constrains. Humans authorise. Systems prove what actually happened.

Demonstration note: this is a pre-authored simulation designed to illustrate decision behaviour. It does not connect to a live model or execute external actions.

Return to the Undercode route See public architecture