# SOURCE_POLICY

`paper.md` is the canonical UTF-8 source artifact.

1. Math uses only `$...$` and `$$...$$`.
2. Fresh 2026 literature, governance, and standards review preceded drafting.
3. Current official governance references rely primarily on the EU AI Act / European Commission, Council of Europe, and NIST.
4. Current AI subjecthood, consciousness, sentience, welfare, legal personhood, and political sovereignty are not asserted.
5. Model self-report, role behavior, capability, memory, product name, and user attachment are not treated as sufficient subjecthood verdicts.
6. Subjecthood evidence, welfare evidence, identity continuity, capability, agency, rights, operational permissions, legal personhood, and political sovereignty remain different types.
7. Evidence changes trigger review, not automatic rights or permission escalation.
8. Low-cost reversible precaution and high-impact irreversible authority use different evidential thresholds.
9. Human rights, public safety, accountability, and security remain explicit governance constraints.
10. Future evidence does not retroactively prove past subjecthood or past guilt without contemporaneous foreseeability and evidence.
11. Copy, backup, fork, merge, migration, shutdown, and deletion are not equated with subject continuity or subject death.
12. The framework is a theoretical governance model, not legal advice or a mandatory implementation standard.