Plain-language summary: Our AI is built to help you save money on idle licenses — not to harvest your data, automate decisions about people, or train external models. Humans approve every action.
Purpose: This document outlines AppReclaim's principles and controls for responsible AI development and deployment.
1. Core Principles
- Beneficial: AI features are designed to help customers save costs and optimize license usage, not to maximize our revenue or engagement
- Transparent: We clearly disclose where AI is used and how it influences recommendations (see AI Transparency)
- Secure & Private: Customer data is protected; AI does not train on or share Customer Data with third parties
- Human-in-the-Loop: All consequential actions require explicit human approval before execution
- Accountable: We maintain audit trails and accept responsibility for AI behavior within our platform
2. Implementation Controls
- Pre-Execution Checks: Validation of AI recommendations before presenting to users
- Post-Execution Verification: Confirmation that executed actions match approved intentions
- Least-Privilege Scopes: OAuth integrations request only necessary permissions
- Grace-Period Restores: Soft-delete periods where platform supports, enabling recovery from unintended actions
- Rate Limiting: Controls to prevent bulk unintended actions
- Rollback Capabilities: Restore paths available for supported platforms; 30-day default window, 60 days on Enterprise tier and inherited managed clients
3. Data Practices
- AI prompts route through the Lovable AI Gateway (currently upstream providers: Google Gemini, OpenAI). Both the Gateway and upstream providers are contractually bound not to train on Customer Data.
- PII is minimized in prompts; we send aggregated usage data and SKU metadata whenever possible.
- Prompt-injection and content-filtering guardrails are applied at the Gateway and at the application layer.
- Rate limiting prevents runaway costs and reduces the blast radius of unintended bulk operations.
- Third-party AI providers are bound by data processing agreements via the Gateway.
4. Bias & Fairness
- AI recommendations are based on objective usage data, not demographic factors
- We regularly review AI outputs for systematic biases
- Customers can override AI recommendations at any time
5. Review & Updates
This policy is reviewed periodically to ensure alignment with:
- Emerging AI regulations and standards (including the EU AI Act and U.S. state AI laws)
- Industry best practices
- Customer feedback and concerns
- New AI capabilities added to the platform
6. Reporting Concerns
If you have concerns about AI behavior in AppReclaim, contact legal@appreclaim.com. We take all reports seriously and investigate promptly. See also the AI error reporting path in AI Transparency §7.
7. Model Lifecycle
We may switch upstream model providers or versions to improve quality, latency, or cost. When we do, we will not regress the safety, privacy, or human-in-the-loop guarantees stated in this policy. Material changes to upstream providers will be reflected in the Sub-Processor list.
8. Change Log
- May 12, 2026 — Named Lovable AI Gateway and current upstream providers; added content filtering / prompt-injection guardrails, EU AI Act mention, and model lifecycle commitment (§7).
- January 18, 2026 — Initial publication.
