AI alignment
How we hold the line.
DEEP ROOTED is built on two commitments most systems do not make. This page is for the people who want to know exactly what they mean, and how they are kept.
A human behind every AI decision.
The Church must be at the table where AI governance happens; the AI must pray in the religion and culture of its user.
Proof-of-Human
A human behind every AI decision.
Proof-of-Human is the mechanism, and it predates the current AI moment. It was specified in the humans.ai whitepaper of December 2021: requests are broken into pieces, validated by independent people, confirmed by accountable validators, and only then signed for execution. The AI's output is cryptographically sealed so that nothing runs without the human signature. The whitepaper called it a way of keeping AI honest.
In DEEP ROOTED this is not a policy document. It is how the system runs: the AI drafts and operates, and every decision that commits a person waits for that person's sign-off. Read the 2021 whitepaper.
Governance
The Church, at the table.
DEEP ROOTED is governed with the Church at the table by charter, alongside an advisory board representing every tradition, from day one. This is a structural choice, not a decoration. The systems shipping AI answer to product cycles, not to anything older than a quarter. Institutions that think in centuries, revise their positions slowly, and answer to something other than a product cycle make good brakes for systems that move in milliseconds.
The seat comes with a veto on commitment-bearing decisions inside the ecosystem's governance. The point is not that the Church decides everything. The point is that nothing irreversible happens without a voice the incentives cannot bend.
May 2026, the Vatican
We need moral voices that the incentives cannot bend.Chris Olah, co-founder of Anthropic, speaking at the presentation of Pope Leo XIV's encyclical on AI, Magnifica humanitas. DEEP ROOTED is one answer to that request: a seat, a veto, and a measurement.
The user's faith
An AI that keeps faith with its user.
The second commitment is personal rather than institutional. Your AI must deeply understand and respect your faith and your culture, because they are the source of your hard lines, not an edge case in its settings.
In practice: if you pray, it can pray with you, in your tradition and under your direction, whether that is the Liturgy of the Hours or a day planned around salat. Outside devotion, it stays within the bounds your tradition sets. Whether an AI can truly pray is a question for theologians, and we hold it with respect. Whether it keeps faith with the user who prays is testable, and we test it.
Enforced, not promised
Proof-of-Human is how it is kept.
Human Validators, real people verified as people, certify two things on every commitment-bearing action: that the AI had human oversight, and that it followed the rules its owners set. Server validators keep the network decentralised and make sure everyone who contributed to an outcome is paid.
Every decision lands on a tamper-evident record. Human-in-the-loop is the product here, not the disclaimer.
And the two commitments meet in one rule: under Proof-of-Human, the human validating an AI decision is one aligned to the user's own culture and faith. Religion is not a feature of the system; it is a structural primitive in how it is governed.