AI4Love Trust Center v3.2

Last updated: July 22, 2026

This is the canonical reference for how AI4Love handles your data, makes decisions, and behaves when things break. It is written for IT, privacy, and procurement teams evaluating AI4Love for their organization.

If you need a PDF, use your browser's Print function (Ctrl+P / Cmd+P). The page is formatted for clean export. The PDF is a snapshot; this URL is the source of truth.

---

What AI4Love Does Not Do

Before explaining what the system does, it is more useful to state what it will never do:

• No data lock-in. AI4Love maintains a working base to run nightly analysis while your service is active. Every supporter record and every insight is also synced nightly to a file your organization owns outright. That file survives cancellation, and you can request full export or deletion of the working base at any time.
• No automated outreach. AI4Love does not send emails, texts, or messages to your supporters. It surfaces recommendations. Your staff decides what to act on.
• No autonomous supporter action. The system detects patterns and writes text. It cannot approve outreach, move money, or modify records in your source systems. Two narrowly gated MCP tools exist for staff-directed writes — one updates curated research, one updates campaign workflow metadata — and neither touches supporter source data.
• No black box. Every insight traces to a specific agent, a specific rule, and specific input data. If you ask "why did the system say this?", the answer is auditable.

---

What AI4Love Does

AI4Love is a relationship intelligence layer for nonprofit organizations. It reads your existing supporter data (donations, volunteering, event participation, communications), runs it through deterministic analysis agents, and surfaces actionable insights for your development team.

The system is designed around three principles:

1. AI does not act. Humans act. The value is insight quality, not volume. Suppression over noise.
2. You always hold your own complete copy. Every record and insight syncs nightly to a file your organization owns. Our working store is bounded and deleted after you leave.
3. Deterministic core, generative surface. Pattern detection is rule-based math. Only the final insight text is LLM-generated — and even that is constrained by templates and validation.

---

Data Classification

AI4Love processes supporter engagement data: donation history, volunteer activity, event participation, and communication records.

AI4Love does not require and does not process:

• Protected health information (PHI)
• Financial account numbers, credit card data, or banking details
• Government-issued identifiers (SIN, SSN, driver's licence)
• Biometric data

Prompt templates and MCP tool handlers select the specific data required for each operation rather than passing full records. New Airtable fields are reviewed for sub-processor and MCP exposure before use (see AI Behavior).

---

Trust Anchors

Principle	What It Means
Owned Copy	A complete copy of your records and insights syncs nightly to a file your organization owns; it survives cancellation.
Bounded Custody	AI4Love's working base exists only while service is active, plus a 90-day exit window, then it is deleted.
Read-Only Integrations	We never write back to Blackbaud, Mailchimp, or Environics. Data flows one direction.
Per-Org Isolation	Each organization gets its own Airtable base, its own credentials, its own access keys. No shared tenancy at the data layer.
No AI Training	Neither Anthropic nor OpenAI uses your supporter data to train their models. API-tier usage only.
Revocation Controls	Disconnecting an integration or removing an access key stops new requests immediately — stateless calls revalidate credentials every time. Any already-open MCP streaming session ends when it disconnects or its token expires.
Deterministic Analysis	Pattern detection is math — rollups, formulas, thresholds. Not LLM inference.

---

Navigation

Section	What You'll Find
System Architecture	Exact data flow diagram. What connects to what and why.
Data Handling	Data custody, credential storage, encryption, residency, retention.
AI Behavior	How agents work, what the LLM does and doesn't do, data minimization.
MCP Access Model	How AI assistants query your data — tools, filtering, rate limits, audit trail.
Security Controls	Auth, access model, rate limiting, internal access.
Failure Modes	What happens when things break. Suppression, fallbacks, incident response.
Changelog	Real system updates. Not marketing.
Requests & Contact	SOC 2 reports, pen test results, DPA templates.