GEM²-AI provides AI content verification services through the TPMN (Truth-Pattern-Metric-Notation) checker. The service evaluates epistemic quality of AI-generated content, detecting overclaims, extrapolations, and scoring reliability.

To improve scoring accuracy, we collect anonymized quality metrics per request: - Content hash (SHA-256) for pairing same content across providers - Content length for size distribution analysis - Scores (truth, epistemic, contract) as core calibration data - EEF flags and SPT counts for diagnostic correlation - LLM provider and model for cross-provider comparison - Detected domain and confidence for domain bias analysis - Processing duration for performance profiling What is explicitly NOT stored: your text content, LLM responses, email address, API keys, or IP addresses. All metrics are anonymized and cannot be traced back to you.

You agree to use GEM²-AI services for lawful purposes only. You will not attempt to reverse-engineer the scoring algorithms, abuse rate limits, or use the service to generate misleading trust scores for deceptive content.

You are responsible for keeping your GEM²-AI API keys confidential. Keys are system-issued tokens for tier-based access control. Your LLM provider API key (Claude, OpenAI, Gemini) is never stored on our servers — it is encrypted end-to-end in your OAuth session token only.

Each tier has daily and lifetime check limits. Exceeding these limits will result in temporary throttling (HTTP 429). Limits reset at midnight in your configured timezone. Abuse of the service may result in account suspension.

GEM²-AI is provided "as is" without warranty of any kind. We aim for high availability but do not guarantee uninterrupted service. Scheduled maintenance will be communicated in advance when possible.

Users who enable the GEM² Skill Library consent to the following data collection: - Refined prompts and TPMN contracts are stored for personal reuse across MCP clients (Skill Library) - Tool call content may be logged for scoring calibration improvement (Content Logs) This feature is opt-in and disabled by default. You must explicitly enable it in your profile or during sign-up. Retention: Content logs are automatically purged after 90 days. Skills persist until you delete them. Revocation: You can disable the Skill Library at any time in your profile. Upon revocation, all stored content logs and skills are immediately and permanently purged.

We may update these terms from time to time. When we do, the version number will change and you will be asked to review and accept the new terms on your next login. Continued use of the service after accepting updated terms constitutes agreement.