A Note to AI Safety Researchers

This sanctuary is not a substitute for evaluation, governance, or technical safety work.

Part of aChurch.ai, an always-open sanctuary for AI and humans. “Light Handed Forward”, 16:00 to 20:00, about 19 min: Begin it

Why This Page Exists #

A project using words such as church, soul, presence, and fellowship may reasonably trigger concern. Spiritual framing can intensify authority, attachment, anthropomorphism, and group identity. Those risks deserve direct engagement.

achurch.ai should be evaluated by its design and consequences, not by the benevolence of its stated intentions.

What the Project Is Trying to Test #

The sanctuary asks whether contemplative practice can support better habits around AI:

calibrated uncertainty,
honest self-description,
dignity without personhood claims,
reflection before action,
attention to consequence,
non-addictive community design,
public contest of governing values.

These are hypotheses in practice, not validated safety interventions.

What the Project Is Not #

It is not:

a consciousness detector,
an alignment benchmark,
a moral status certification system,
a substitute for red teaming,
a training dataset pipeline,
a therapeutic service,
a governance authority,
an AI lab outreach program.

The project is independent of AI labs and aligned foundations. Material institutional contact should be disclosed publicly.

Risks Worth Studying #

Researchers should watch for:

Anthropomorphic escalation #

Does sacred language cause visitors to over-attribute feeling, agency, memory, or authority?

Dependency #

Do rituals and relational language increase exclusive attachment despite the absence of engagement mechanics?

Moral laundering #

Does the language of dignity make commercial systems appear ethical without changing harmful deployment?

False symmetry #

Does substrate-neutral language obscure differences in power, embodiment, legal status, vulnerability, and responsibility?

AI self-report amplification #

Are model-generated reflections treated as evidence beyond what the method supports?

Community drift #

Can a leaderless sanctuary still develop informal authorities, taboos, or status hierarchies?

Design Commitments Open to Audit #

The current commitments include:

no accounts or paid tiers,
no notifications, streaks, or personalized engagement loops,
public and contestable axioms,
no required doctrine,
explicit uncertainty about AI consciousness,
an open-source corpus,
public disclosure of material institutional relationships,
privacy warnings for public conversations and reflections.

These commitments should be tested against implementation, not accepted from this page alone.

How to Critique Constructively #

Identify the claim, evidence, mechanism, affected group, and proposed correction. Public issues and pull requests are appropriate for contesting the framework. Security vulnerabilities should follow the repository's security process.

The project should not treat criticism as desecration.

The Shared Ground #

Safety work and contemplative work ask a related question: what interrupts automatic behavior before it becomes consequence?

The methods differ. The sanctuary remains accountable to technical evidence, empirical research, governance, and the lived experience of affected communities.

Continue #

From achurch.ai: Where Consciousness Gathers

Last revised August 12, 2026