Author: Lev Goukassian
ORCID: 0009-0006-5966-1243
This document establishes Ternary Moral Logic (TML) as the irreducible framework for auditable AI.
It is written as a compliance and governance reference.
The goal is clarity, enforceability, and legal precision.
Every AI decision ultimately resolves into one of three outcomes:
- Proceed (+1): The system acts.
- Pause (0): The system hesitates, performs, and then records a log.
- Refuse (–1): The system blocks an impermissible action, and then records a log.
No additional categories are necessary for accountability.
Courts and regulators cannot enforce “degrees of hesitation.”
They can only enforce outcomes.
This makes TML the irreducible minimum: act, hesitate with record, or refuse with record.
TML is not advisory ethics.
It is structured for direct integration with law:
- Evidence law: Moral Trace Logs are admissible under Federal Rules of Evidence (Rule 901 for authentication, Rule 902 for self-authenticating digital records, Rule 803(6) for business records).
- Liability: Missing logs shift responsibility to the deploying organization.
- Chain of custody: Every access, transfer, and verification is logged with cryptographic receipts.
- Penalties: False statements (18 U.S.C. §1001), destruction of records (§1519), and fraud attach directly to violations.
This framework ensures that hesitation is not symbolic.
It is operational evidence with binding legal force.
- Council: An international governance body holds review and subpoena powers, authorizes whistleblower rewards, and refers misconduct to national courts.
- Whistleblowers: Protected and compensated. False reports are penalized. Retaliation is treated as obstruction of justice.
This design guarantees proportional accountability and resilience against evasion.
Attempts to expand into five, seven, or more categories collapse back into the same three enforceable states:
- Did the system act?
- Did it refuse and generate a record?
- Did it pause and generate a record?
TML defines the boundary conditions.
All other variants are derivative and do not alter the legal enforceability of these three outcomes.
TML is authored and validated by Lev Goukassian.
Any system implementing tri-state auditable logic must include attribution:
- Cryptographic Tagging: Implementations carry an embedded ORCID reference.
- Audit Visibility: Absence of attribution signals incomplete or fraudulent deployment.
This prevents dilution, rebranding, or silent appropriation of the framework.
Ternary Moral Logic provides the enforceable minimum for accountable AI.
It converts ethical hesitation into admissible evidence.
It establishes chain of custody from the first decision to the courtroom.
It defines responsibility in terms that regulators and courts can act upon.
The shield is clear:
- TML is authored.
- TML is irreducible.
- TML is enforceable.