OSS violation detection

Does AI Guardrails's OSS violation detection support configuring multiple webhook addresses?

The incremental scan for OSS violation detection currently supports only one task. Therefore, you cannot configure multiple webhook addresses. The full scan feature supports configuring a webhook address for each task. This lets you configure multiple addresses.

Can I use OSS violation detection if I have not uploaded images to OSS?

No, you cannot. OSS violation detection applies only to image and video files in OSS buckets. After you specify the OSS buckets to scan in the AI Guardrails console, the images and videos in those buckets are scanned.

Which file formats does OSS violation detection support? Why is a file in my OSS bucket not scanned even though its format is supported?

For more information about the file formats that OSS violation detection supports, see Introduction to OSS violation detection. Files in unsupported formats are not scanned.

If your file is in a supported format but does not have the correct file extension, such as .mp4, it cannot be scanned.

Can OSS violation detection scan historical images?

Yes, it can. You can use the full scan feature to scan historical images in a specified OSS bucket.

What action does Alibaba Cloud take when pornographic images are found by OSS violation detection?

Alibaba Cloud takes no action. The AI Guardrails service only scans content and provides the detection results to you. The service does not view or process your data in any way.

Why can a non-compliant image still be accessed from the internet after it is frozen?

For these cases, you can set the freeze method to Remove non-compliant file in the OSS violation detection settings.

The image might also be accessible because it is cached by CDN. You can configure auto-refresh for the CDN cache. For more information, see Attach a CDN-accelerated domain name.

Which regions does OSS violation detection support?

OSS violation detection currently supports scanning OSS buckets in the following regions:

• oss-cn-hangzhou.aliyuncs.com is the endpoint for an OSS bucket in the China (Hangzhou) region.

• oss-cn-shanghai.aliyuncs.com: The endpoint for an OSS bucket in the China (Shanghai) region.

• oss-cn-qingdao.aliyuncs.com: The endpoint for OSS in the China (Qingdao) region.

• oss-cn-beijing.aliyuncs.com: The endpoint for an OSS bucket in the China (Beijing) region.

• oss-cn-shenzhen.aliyuncs.com: The endpoint for an OSS bucket in the China (Shenzhen) region.

Does AI Guardrails support scanning the URLs of images and videos in OSS?

Yes, it does. We recommend that you use the one-click OSS violation detection feature to scan the URLs of images and videos in OSS.

If you want to use the AI Guardrails API to scan images and videos in OSS, you must use URLs that are accessible over the internet.

Why do all OSS content moderation results show that images failed to load? Does OSS content moderation support callbacks?

This can happen if hotlink protection is enabled for the OSS bucket that you want to scan. Hotlink protection restricts access to resources in your bucket by checking the Referer header. Access is granted only to domain names in the Referer whitelist or if an empty Referer is allowed. If you have enabled hotlink protection for OSS, add the domain name http://yundun.console.aliyun-inc.com to the Referer whitelist.

For more information, see Hotlink protection.

Does OSS violation detection support callbacks?

AI Guardrails can send the results of machine content detection and human review through asynchronous message notifications. If your business requires this feature, you must configure the notification service.

For more information about how to configure callback messages, see Configure message notifications.

How long are the results of OSS violation detection stored?

The AI Guardrails console stores detection results from full and incremental OSS scans for seven days. We recommend that you export the scan results promptly. You can also receive detection results through asynchronous message notifications.

For more information, see View scan results and Configure message notifications.

Can I configure different webhook addresses for different buckets in AI Guardrails's OSS violation detection?

This depends on the scan type. For incremental scans, you can configure only one webhook address for all buckets. For full scans, you can create a separate task for each bucket and configure a different webhook address for each task.

For more information, see Create a full scan task.

If I increase the freeze threshold for OSS violation detection and rescan, can previously frozen images be restored?

OSS violation detection does not currently support automatic restoration. To restore frozen files, you can use the OSS access control API to perform permission operations.

For more information, see PutObjectACL.

If an image's storage class is changed from Archive Storage to Standard for an incremental OSS scan, will the modified image be scanned?

Yes. You cannot directly change the storage class of an object from Archive Storage to Standard. This change requires you to read the object and then rewrite it to the bucket. This rewrite operation is treated as a new upload, which triggers the incremental scan. As a result, the modified image is scanned.

Are there any requirements for the storage class of OSS objects for AI Guardrails's OSS violation detection?

There are no limits for the Standard, Infrequent Access, and Archive Storage classes.

Why does the AI Guardrails console for OSS violation detection show that an image failed to load?

This usually happens because the original image at the specified address was deleted or has expired.