Submission 546
Three Primitives Research Lab
Submitter group: Research institutes.
An AI model identified the submitter group and wrote the summary, key points and recommendation summaries. Counts do not measure importance or merit.
Summary
Read the original submission for the full text.
Argues that its mathematical proof rules out AI supplying authority for decisions about people. Says incident review and reporting occur after action, while website hardening shifts the burden to affected parties. Its attached papers argue for declared human authority before execution and describe the lab’s own safety system. The white-paper author says they are not a lawyer and presents an evidence brief.
Key points
Quotes are the submitter’s own words. Text taken from documents, including scanned pages read by text-recognition software, can differ from the PDF. Check the original. Listed in the same order as on the submission’s page in the interactive guide.
Says a mathematical proof it published shows that no AI system can supply the authority for a decision about a person.
There is a mathematical proof that no AI system can supply the authority for a decision about a person. It was published on 10 January 2026 as FR02, AI Cannot Govern AI, and rests on FR01, the Canonical Logic Sequence, published 28 December 2025.
Says it names the failure mode Ghost Authority, in which a decision takes effect with no human who declared it, and that its lemma proves this cannot be closed from inside the system that produced it.
We named the failure mode Ghost Authority: a decision takes effect with no human who declared it. FR05, the Ghost Authority Lemma, proves it cannot be closed from inside the system that produced it.
Argues incident review and mandatory reporting follow an act, while website hardening shifts burdens to affected parties; none supplies permission for the next action.
Compute spent reviewing incidents is detection after the act. A mandatory reporting regime is notification after the act. Hardening the websites moves the burden onto the party that was acted upon. None of the three produces permission for the next action, because none of them can.
Argues that AI safety guardrail products and verifiers share one structure, in which a computed pass is treated as permission to act, with no declared human authority required at the gate.
The engineering differs. The structure does not; a computation produces a pass, and the pass is treated as the permission to act. No product in this landscape requires a declared human authority at the gate as a structural condition of operation.
Argues that if staff approving an I-CAN plan cannot meaningfully depart from its computed budget, approval is not a decision.
If a staff member approving an I-CAN plan cannot meaningfully depart from the computed budget, the approval is not a decision. It is a costume.
Proposes an immutable pre-execution declaration naming purpose, live human authority and constraints, with a record for each gated action.
a declaration, δ, naming the bounded purpose, the live human authority, and the constraints, issued before execution and immutable for that event, with every gated action leaving a record
Recommendations
Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.
No specific recommendations were identified.