Commons AI Future Program
External review entry point
This page organizes the Commons Initiative AI-governance research for external technical review, including the diligence priorities identified by Dr. Thomas Ainsworth: authority, revocation, auditability, meaningful human control, and accountable organizational structure.
This is not an endorsement by Dr. Ainsworth or the Human-Centered AI Governance Institute. The Commons Initiative’s business-registration, domain-infrastructure, and organizational-accountability gates remain open and are tracked separately in the organizational readiness meter.
Central proposition
The program asks whether powerful AI can remain useful while every consequential capability is:
- explicitly granted and narrowly scoped;
- time-bounded and immediately revocable;
- unable to expand its own authority;
- forced into a safer mode when authorization or integrity fails;
- recorded in a tamper-evident audit trail;
- subject to meaningful human confirmation at irreversible decision points; and
- allowed to retain only bounded, reviewable lessons from prior decisions.
The present contribution is a governance architecture and test program. It is not evidence of machine consciousness, quantum communication, temporal control, or a new law of physics.
Six-part evidence package
- Canonical volume index
- Duplicate and missing-volume audit
- Claim-evidence ledger
- Corrected engineering edition
- Speculative research edition
- Implementation and test program
Working software
The repository now includes a small reference implementation in commons_governance/ and tests in tests/test_commons_governance.py. The software covers:
- scoped consent grants;
- expiry and revocation;
- global freeze and reviewed restoration;
- hash-chained audit events;
- bounded moral-memory records; and
- fail-closed authorization decisions.
Run the reference checks with:
python -m unittest discover -s tests -p "test_*.py"
What a reviewer can do
- Inspect the claim-evidence ledger.
- Reproduce the software tests.
- Attempt unauthorized scope expansion, replay, expiry bypass, ledger tampering, and restoration without review.
- Identify the most important missing control.
- Treat every failed test as a preserved research result.