Skip to main content

​
Applicability

This policy applies to the ACG API Relay. This document does not apply to “Wbot”, “Womni”, or other W-series models. This policy is provided in both Chinese and English. The English version is an AI translation and is for reference only. In the event of any discrepancy or dispute, the Chinese version shall prevail.

​
1. Data submitted to and returned by our inference services

We understand that the input you provide to our API and the output generated by the API may contain your personal information. Unless we obtain your explicit consent, we will not store, sell, or use this data for training. We may sometimes store API call inputs and outputs for a limited time for debugging purposes.

​
2. Data privacy during inference

​
2.1 Self-deployed models

Definition: Models labeled as “self-deployed models” by the WZH Team. When using self-deployed models on the ACG API Relay, you can be assured that your data is safe. We do not store the data you submit to the API on disk. We only store it in memory during inference. After inference is complete, the data is deleted from memory. We also do not store the output of the inference process. After inference is complete, the output is returned to you and then deleted from memory. When using image generation models, the output will be stored by us for easy access for a short period of time.

​
2.2 Non-self-deployed models (external models)

Definition: Models whose type is not explicitly stated. When using external models on the ACG API Relay, we will do our best to maintain the security of your data. We require upstream service providers to handle the data you submit to the API as follows:
  1. Not use your submitted data for model training or similar operations.
  2. Adjust the retention period of your submitted data to the minimum time necessary to achieve the processing purpose.
  3. Not share the data you submit to the API with any non-compliant third parties.
After inference is complete, the output may be stored by the upstream service provider for up to 31 days. When using image generation models, the output may be stored by the upstream service provider. The retention period is typically 30 to 90 days. When the data you submit to the API contains content that is not allowed by the upstream service provider, your submitted data may be stored for review by relevant personnel.

​
2.3 free_c group models

Definition: Models within the free_c group. When using free_c group models on the ACG API Relay, we will do our best to maintain your data privacy. However, as with most public welfare services, it is unavoidable that the data you submit to the API may be processed by service providers as follows:
  1. Use your submitted data for model training or similar operations.
  2. Store your submitted data indefinitely.
  3. Share data with unknown third parties.
This is nearly impossible to avoid. However, unlike most public welfare services, to protect your data privacy as much as possible, we use the WDN content firewall to filter the data you submit to the API (only for free_c group models). Our firewall will filter out metadata that could potentially identify you. The WDN firewall’s data usage policy can be referenced from the “Self-deployed models” section above.