AI Capability and Algorithm Disclosure
This page explains where the DeepSData website uses artificial intelligence, where the underlying algorithm comes from, how your input is handled, and how to request a review of AI-assisted results.
1. Where AI is used on this website
Large language model technology is used in exactly two features:
(a) Requirement assistant — a conversational helper that turns your data needs into a structured requirement confirmation sheet;
(b) Data availability assessment report — based on the requirements you confirm, the system runs real searches across authoritative data platforms and public sources; the language model checks each candidate source against your stated conditions item by item, and the final report is assembled by system code following fixed rules.
AI on this website is limited to the two scenarios above. We do not offer public-facing generation of text, images, audio or video as deep-synthesis content, we do not generate news content, and we do not provide internet news information services.
2. Algorithm source
The language model capability is accessed through the API of an established third-party model provider that has completed the security assessment and algorithm filing required under applicable Chinese regulations. DeepSData has not developed its own large language model algorithm, and we have a service agreement with the provider covering service scope, data security and compliance responsibilities.
3. How your input is handled
What you type during conversations and searches is used only to produce that answer or that assessment report — nothing else. Search keywords are scrubbed for sensitive information before being sent to external public data platforms. We do not request or accept sensitive personal information such as ID numbers, bank accounts or medical records, and the system automatically blocks such input. For full details on personal information, see our Privacy Policy.
4. Accuracy limits and how to request a review
AI-assisted content can contain errors, so we apply a cautious standard: whether a candidate dataset counts as a "match" is decided by system code checking each of your must-have conditions one by one — we do not take the model's own claims at face value. A search cannot guarantee that a fully matching dataset exists; anything that falls short is stated plainly in the report. If you dispute any content in a report, email contact@deepsdata.com and we will review it and reply promptly.
Operator: DeepSData (Shenzhen) Data Intelligence Co., Ltd. (Unified Social Credit Code: 91440300MAKC8NXM0D)
Email: contact@deepsdata.com
Last updated: July 2026