AI-use registry

Where AI is used

AI assists — never decides — in: grant discovery (finding opportunities), fit scoring (ranking prospects), application drafting, extracting a grant from a URL, report drafting, intelligence insights, resource localization and Q&A, organization-profile autofill from a website, the Ask WeShield assistant, and drafting in the app builder.

Every one of these is assistive. AI may propose a draft, a score, or a discovery; it cannot send, publish, submit, or connect anything.

The human gate

A human principal reviews and ratifies every AI output before it has effect. This is enforced in the database, not just the interface: AI agents are blocked at the data layer from holding the ratify, publish, or connector capabilities — they cannot bypass review even with direct access.

What the model receives

Prompts are derived from your workspace data only for the feature you invoke. High-risk identifiers (such as emails, phone numbers, card numbers, and government IDs) are redacted before any prompt leaves the platform. The active inference provider is disclosed on the sub-processor list and may process data outside Canada; AI features are opt-in per workspace.

Accountability and cost control

Every AI call is recorded on the tenant’s tamper-evident audit trail. AI spend is metered against a platform monthly ceiling with alerting and an administrator kill-switch, so cost cannot run away unobserved.

Training

WeShield does not use your workspace data to train models, and does not sell it. Whether a third-party inference provider may use submitted content is governed by that provider’s own terms and depends on which provider is active: our supported United States provider (Anthropic) does not train on data submitted through its API, while other providers’ terms vary. The active provider and its data-use posture are disclosed on the sub-processor list, and AI features are opt-in per workspace.