AI Data & Model Privacy Policy

1. Architectural Philosophy: AI as a Compiler

At Rayvotech Tools Inc. ("Rayvotech", "we", "us", or "our"), our engineering philosophy treats artificial intelligence as a compiler pass rather than a content owner. When you build a web application on Rayvotech Tools, our orchestrator transforms natural language directives into formal Abstract Syntax Trees (ASTs), validates component hierarchies, generates relational SQL schemas, and emits standard Vite/React source code.

This document establishes our binding commitments regarding the confidentiality, security, and intellectual property rights associated with AI-assisted software synthesis.

2. Zero Model Training Guarantee

Contractual Commitment: Rayvotech Tools legally guarantees that your proprietary prompts, uploaded assets, custom code refactors, database table structures, and customer records are NEVER used to train, retrain, fine-tune, or evaluate public foundation AI models.

We do not license, sell, or provide customer codebases or prompt histories to any third-party AI laboratory or data broker for model training purposes under any circumstance.

3. Enterprise Multi-Model Routing Pipeline

Rayvotech Tools leverages specialized frontier intelligence models (including GPT-6 Astra, Claude Opus 5.5, and Gemini 3.8 Flash) through dedicated enterprise API endpoints.

Context Isolation: Requests transmitted to inference providers contain only the minimal AST sub-tree and type definitions required to satisfy the specific refactor or compilation instruction. Prompts are transmitted over encrypted TLS 1.3 tunnels and processed in volatile memory.

4. Zero-Data-Retention (ZDR) Commercial Terms

All upstream foundation model providers utilized by Rayvotech Tools operate under commercial enterprise agreements that include enforceable Zero-Data-Retention (ZDR) guarantees:

• Inference providers do not log or store prompt payloads or generated completion tokens to persistent disk.
• Inference providers are contractually prohibited from using API requests for model alignment, RLHF (Reinforcement Learning from Human Feedback), or training data curation.
• All input and output tokens are permanently discarded from inference buffers immediately upon stream completion.

5. Intellectual Property Rights in Synthesized Code

You Own All Outputs: To the maximum extent permitted by applicable copyright law, you hold 100% full ownership, title, and intellectual property rights in all source code, layout structures, stylesheets, and database schemas synthesized by the Platform on your behalf.

Rayvotech Tools makes no claims of copyright or co-authorship over the software you generate, nor do we demand ongoing royalties, platform fees, or attribution requirements on exported code.

6. Verification, Testing & Human-in-the-Loop

Probabilistic Nature of AI: While Rayvotech implements automated syntax linting and type checking, deep learning inference is inherently probabilistic. Synthesized code may occasionally contain logic anomalies, edge-case limitations, or deprecated package calls.

Developer Responsibility: You are responsible for reviewing, testing in the live WebContainer sandbox, and validating the security and regulatory compliance of your software prior to deploying to mission-critical business environments or processing live customer payments.

7. Automated Security Passes & Package Linting

Prior to mounting code into in-browser sandboxes, our compilation engine runs automated static analysis passes designed to:

• Prevent the emission of hardcoded secrets or API tokens.
• Verify that relational PostgreSQL database queries enforce Row-Level Security (RLS) policies.
• Validate npm package dependencies against public vulnerability registries to block compromised or malicious libraries.

8. Ethical Guardrails & Prohibited Syntheses

Rayvotech Tools implements automated guardrails to intercept prompts requesting the creation of malicious software (such as phishing kits, credential interceptors, malware, or exploit tooling). Attempts to bypass these safety classifiers will trigger workspace review and termination under our Acceptable Use Policy.