TRUST ARCHITECTURE
Privacy and security are part of human agency.
Attention Inc. is built for trust. Our design commitments ensure that the architecture of the platform honors the very people it serves.
Design Commitments
Human agency requires wise, proactive architecture.
- Minimal data collection
- Clear and informed consent
- Export and deletion rights
- No raw reflection storage
- No training on user content
- Scoped AI functions
- Human approval gates
- Privacy-preserving logs
- Security guidance and training
- AI liability awareness
Human Authority
A person decides. Always.
No artifact we help produce is self-executing. Every workflow we describe identifies a human owner, a verification step, and an approval gate before an output leaves the room. Where work touches people, money, reputation, mission, or safety, the decision belongs to an accountable person, not to a tool and not to us.
Authority also means the authority to stop. Escalation paths, stoppage rights, and rollback steps are written down before a workflow changes, not after something goes wrong.
Evidence and correction
An AI output is a draft until it has been checked against a source of truth. We ask what must be verified, by whom, and against what. That standard applies to our own work as well: if something we produced with you is wrong, incomplete, or misread your context, say so and we will correct it.
Participants can correct or revise their own reflection and prompts. Those corrections stay with the participant.
Private records and organization-owned records
Private to the participant
Reflection, quotations, individual answers, corrections, and revised prompts written inside The Prompt Literacy Assessment. An employer, sponsor, or manager does not receive them. With separate consent, a small set of non-identifying prompt-pattern signals may contribute to an aggregate cohort report, and no report is generated for fewer than ten consented completions.
Owned by the organization
Workflow descriptions, the organizational AI Contribution Map produced inside an organizational engagement, governance and security requirements, control points, implementation plans, and executive readouts. These describe work, roles, tools, data paths, and conditions. They do not contain private participant reflections or employer-facing profiles of individuals.
An individual who enrolls in Faithful Presence on their own keeps their personalized Map. It is not shared with an employer or sponsor. In both cases, the method, SHIELD architecture, tier framework, and templates remain Attention Inc. intellectual property, licensed for use rather than assigned.
Restricted information
Some information should not enter an AI system at all. Classified, regulated, privileged, pastorally confidential, and personally sensitive material is named and bounded before any tool is chosen. Part of the work is deciding where AI should not be used.
We do not ask for your credentials, system prompts, model configurations, or environment values, and we do not expose our own.
A DESIGN COMMITMENT
Privacy and security by design describes how we build and how we advise. It is not a certification, an audit result, a penetration test, or a guarantee of security or regulatory compliance. Read the Privacy Notice for current practice, Responsible Use and Claims for scope, and Accessibility for our current target.