Thrive AI privacy

What we collect when you use Thrive AI

A Thrive AI conversation generates a small, well-defined data set:

• Your messages. What you type, as typed. Encrypted in transit (TLS 1.2+) and at rest (AES-256).
• The model responses. What Thrive AI replied. Stored alongside your messages so the conversation reads back coherently.
• Timestamps. When each message was sent. Used so you can scroll a conversation in order, and so the mood-pattern view across the week works.
• Mood tags (only if you add them). If you use the daily check-in feature and rate your mood, that rating is stored as a number scoped to that day.
• Guardrail events. When the safety layer fires (crisis language, self-harm, harm to others, medical-emergency language), we log that it fired and what category it was in. We do not log the surrounding message text to a separate guardrail-event store; the message itself is already in your account.
• Technical metadata. Account ID, app version, device platform (iOS or Android). This is the minimum needed to deliver the conversation to the right device and to investigate bugs.

Model training, plainly

Saalvio does not train models on your Thrive AI conversations.

We run Thrive AI on top of a foundation model provided by an upstream AI vendor. We pass each of your messages plus the prior conversation history to that model under a contractual data-processing agreement that prohibits the vendor from training on the prompts and responses. The vendor processes each request, returns the response, and discards the prompt under their zero-retention or short-retention terms.

We do use de-identified, aggregated patterns to evaluate Thrive AI quality (for example, how often a guardrail fired, average conversation length, how often a suggested exercise was accepted). Aggregated patterns are different from training data; they describe the system's behaviour without retaining any individual conversation as a training example.

If Saalvio's data-processing terms with our AI vendor change, this page will be updated and the change will be summarised in our changelog.

How long we keep your conversations

Active accounts

Indefinitely, while your Saalvio account is active and you have not deleted the conversation. The conversation history is part of the product (it is how Thrive AI picks up where you left off and how mood patterns work across weeks).

Conversations you delete

Removed from your active account within 24 hours and purged from backups within 30 days. After 30 days they are gone.

Accounts you close

Personal data and Thrive AI conversations are deleted within 30 days of account closure (immediate from production systems, then purged from backups as backup rotation runs).

Inactive accounts

If you have not opened the app in 24 months, we email you and ask whether to keep the account. If you do not reply within 30 days of that email, the account is closed and the data is deleted on the same 30-day schedule.

Aggregated quality data

De-identified patterns about Thrive AI behaviour (guardrail fire rates, average exercise completion rates) are kept indefinitely. They contain no message text and no link back to your account.

Compliance posture by jurisdiction

Thrive AI is operated by Saalvio (Zen Mind Inc., incorporated in Ontario). The data-handling commitments above hold regardless of where you live. The legal framework that backs them varies by country.

Canada

PIPEDA, the federal private-sector privacy law, applies to Saalvio's collection, use, and disclosure of personal information. Where Thrive AI conversation content discusses health, Ontario's PHIPA framework informs our handling even though Thrive AI is not a regulated health-information custodian. Quebec's Law 25 is under evaluation; Thrive AI is technically available in Quebec but Saalvio has not yet completed a formal Law 25 compliance assessment. If you are in Quebec and want certainty, wait for our Quebec readiness milestone.

United States

Thrive AI is not a HIPAA-covered service. Saalvio is not your healthcare provider when you use Thrive AI; you are using a wellness chatbot, not a medical service. We follow HIPAA-aligned administrative and technical safeguards as a baseline (access controls, encryption in transit and at rest, audit logging) without claiming HIPAA-covered status. State-level privacy laws (California CCPA/CPRA, Colorado, Connecticut, Virginia, Utah, and others) apply to the extent Saalvio is processing your personal information; our compliance approach is to apply the strictest state floor to every US user.

Outside Canada and the US

Thrive AI is marketed for Canada and the United States in Phase 1. We do not actively block accounts created from elsewhere, but our compliance posture (GDPR for the EU and UK, for example) is not yet formally certified. We will update this section as we expand.