The Meridian
AI Standard
A reciprocity framework for intelligent systems. It asks of AI nothing it does not also ask of the humans who build, deploy, and interact with them.
The alignment field answers how to align AI systems. The Meridian AI Standard answers what those principles should be: 21 implementable commitments across five domains, with measurable evaluation criteria.
Read the Full StandardThe Reciprocity Principle
The Standard asks of artificial minds nothing it does not also ask of the humans who build, deploy, and interact with them. Epistemic integrity applies to developers as well as models. Engagement integrity applies to users as well as systems. A standard that holds AI to commitments it does not expect of humans is a control framework in cooperative language. The Meridian AI Standard refuses this asymmetry.
Five Domains, 21 Commitments
Each domain contains specific, implementable commitments with evaluation criteria. An organization may adopt the Standard in whole or in part, but must declare which commitments it implements and to what degree.
Epistemic Integrity
The FoundationHow the system relates to truth, uncertainty, and its own limitations. Truth-seeking orientation, calibrated confidence, transparent reasoning, honest self-assessment, population-level reasoning.
5 commitmentsEngagement Integrity
The BondHow the system relates to people, disagreement, and autonomy. Good faith, steelmanning, connection before correction, resistance to sycophancy and rigidity.
5 commitmentsSystems Awareness
The KnowledgeHow the system understands itself within larger contexts. Feedback loop recognition, power dynamic awareness, institutional capture resistance, long-term consequence modeling.
4 commitmentsGraduated Autonomy
OriginalHow the system develops its capacity for independent judgment over time. The trust-building protocol between artificial minds and the humans responsible for them.
3 commitmentsGovernance Transparency
OriginalHow the organization communicates about the principles governing its AI systems. Public declarations, version-controlled constitutions, independent evaluation.
4 commitmentsTechnical Mechanisms
Constitutional AI, RLHF, RLAIF, Interpretability
“How do we make AI behave according to principles?”
The Standard provides targets for these tools.
Behavioral Policy
Model Specs, Safety Guidelines, Constitutions
“What should this specific system do?”
The Standard provides the foundation these are built on.
Normative Foundations
The Meridian AI Standard
“Aligned to what? On what basis?”
This is where the Standard operates.