AI & Data Use Statement

1. Purpose of this statement

The most common procurement questions in the AI era are: "Will our data be used to train AI?" "Which models receive our data?" "Can we avoid overseas models?" This statement answers them in one place, supplementing the Privacy Policy and User Agreement for AI scenarios.

2. Our training-data commitments

(1) Your business data (brand information, knowledge-base documents, content assets, monitoring data, lead data) is never used by us to train any AI model; (2) we do not provide your data to model providers for training their foundation models — our agreements include no-training/no-improvement clauses (with zero-retention APIs where the provider supports them); (3) any in-house model fine-tuning uses only anonymized own data or public data, never customer data.

3. What is sent, and to whom

When you use monitoring or generation, the Platform sends the minimum data required for that task to the relevant model/engine: monitoring — question terms, necessary context and retrieval requirements; content generation — retrieved knowledge-base excerpts and writing requirements; image generation — the prompt. We never send entire knowledge bases or customer information unrelated to the task; sending records are logged (engine call ledger).

4. Model routing and region choice

The Platform connects both domestic and overseas engines (list in Sub-processors). You can: (1) lock AI calls to domestic models only (effective immediately; nothing is sent abroad afterwards); (2) enable/disable overseas engines per function (e.g. overseas for monitoring, domestic for generation); (3) apply a tenant-wide policy (enterprise management scenarios). Cross-border transfers follow section 13 of the Privacy Policy and the DPA.

5. AI content labeling and responsibility

AI-generated content carries explicit labels ("AI generated") and implicit metadata per the Measures for Labeling AI-Generated Synthetic Content. Generated content passes compliance pre-checks (sensitive terms, industry rules, brand guidelines) before publishing, but final compliance responsibility rests with the Customer (User Agreement section 5).

6. Human and automated boundaries

(1) Monitoring data and scoring outputs are deterministic algorithms (recomputable, auditable) — not "model improvisation"; (2) AI-assisted judgment steps (content generation, quality assessment, compliance review support) all retain human review; (3) we do not use customer data for automated decisions (such as auto-pricing or credit scoring); final decisions on content approval and lead grading belong to the Customer.

7. Withdrawal and audit

You may adjust model-use policies at any time; after revoking overseas-model authorization we stop the corresponding transfers within 5 business days and record it in the ledger. Enterprise customers may audit our engine call records (audit rights per DPA section 8).