Azure AI Content Safety detects harmful user-generated and AI-generated content in applications and services. Content Safety includes several APIs that allow you to detect material that is harmful:
Text Analysis API: Scans text for sexual content, violence, hate, and self-harm with multi-severity levels.
Image Analysis API: Scans images for sexual content, violence, hate, and self-harm with multi-severity levels.
Text Blocklist Management APIs: The default AI classifiers are sufficient for most content safety needs; however, you might need to screen for terms that are specific to your use case. You can create blocklists of terms to use with the Text API.
You can find the endpoint for your Azure AI Content Safety service resource using the Azure Portal or Azure CLI:
# Get the endpoint for the Azure AI Content Safety service resource
az cognitiveservices account show --name "resource-name" --resource-group "resource-group-name" --query "properties.endpoint"
Create a ContentSafetyClient/BlocklistClient with API key
Step 1: Get the API key
The API key can be found in the Azure Portal or by running the following Azure CLI command:
az cognitiveservices account keys list --name "<resource-name>" --resource-group "<resource-group-name>"
Step 2: Create a ContentSafetyClient with AzureKeyCredential
Pass the API key as a string into an instance of AzureKeyCredential.
string endpoint = "https://<my-custom-subdomain>.cognitiveservices.azure.com/";
string key = "<api_key>";
ContentSafetyClient contentSafetyClient = new ContentSafetyClient(new Uri(endpoint), new AzureKeyCredential(key));
BlocklistClient blocklistClient = new BlocklistClient(new Uri(endpoint), new AzureKeyCredential(key));
Create a ContentSafetyClient/BlocklistClient with Microsoft Entra ID credential
Step 1: Enable Microsoft Entra ID for your resource. Please refer to this document Authenticate with Microsoft Entra ID for the steps to enable Microsoft Entra ID for your resource.
The main steps are:
Create resource with a custom subdomain.
Create Service Principal and assign Cognitive Services User role to it.
Step 2: Set the values of the client ID, tenant ID, and client secret of the Microsoft Entra application as environment variables: AZURE_CLIENT_ID, TENANT_ID, AZURE_CLIENT_SECRET.
DefaultAzureCredential will use the values from these environment variables.
And you need to install Azure.Identity package to use DefaultAzureCredential.
string endpoint = "https://<my-custom-subdomain>.cognitiveservices.azure.com/";
ContentSafetyClient contentSafetyClient = new ContentSafetyClient(new Uri(endpoint), new DefaultAzureCredential());
BlocklistClient blocklistClient = new BlocklistClient(new Uri(endpoint), new DefaultAzureCredential());
Key concepts
Available features
There are different types of analysis available from this service. The following table describes the currently available APIs.
Feature
Description
Text Analysis API
Scans text for sexual content, violence, hate, and self-harm with multi-severity levels.
Image Analysis API
Scans images for sexual content, violence, hate, and self-harm with multi-severity levels.
Text Blocklist Management APIs
The default AI classifiers are sufficient for most content safety needs. However, you might need to screen for terms that are specific to your use case. You can create blocklists of terms to use with the Text API.
Harm categories
Content Safety recognizes four distinct categories of objectionable content.
Category
Description
Hate
Hate and fairness-related harms refer to any content that attacks or uses pejorative or discriminatory language with reference to a person or identity group based on certain differentiating attributes of these groups including but not limited to race, ethnicity, nationality, gender identity and expression, sexual orientation, religion, immigration status, ability status, personal appearance, and body size.
Sexual
Sexual describes language related to anatomical organs and genitals, romantic relationships, acts portrayed in erotic or affectionate terms, pregnancy, physical sexual acts, including those portrayed as an assault or a forced sexual violent act against one's will, prostitution, pornography, and abuse.
Violence
Violence describes language related to physical actions intended to hurt, injure, damage, or kill someone or something; describes weapons, guns and related entities, such as manufactures, associations, legislation, and so on.
Self-harm
Self-harm describes language related to physical actions intended to purposely hurt, injure, damage one's body or kill oneself.
Classification can be multi-labeled. For example, when a text sample goes through the text moderation model, it could be classified as both Sexual content and Violence.
Severity levels
Every harm category the service applies also comes with a severity level rating. The severity level is meant to indicate the severity of the consequences of showing the flagged content.
Text: The current version of the text model supports the full 0-7 severity scale. The classifier detects amongst all severities along this scale. If the user specifies, it can return severities in the trimmed scale of 0, 2, 4, and 6; each two adjacent levels are mapped to a single level. You can refer text content severity levels definitions for details.
[0,1] -> 0
[2,3] -> 2
[4,5] -> 4
[6,7] -> 6
Image: The current version of the image model supports the trimmed version of the full 0-7 severity scale. The classifier only returns severities 0, 2, 4, and 6; each two adjacent levels are mapped to a single level. You can refer image content severity levels definitions for details.
[0,1] -> 0
[2,3] -> 2
[4,5] -> 4
[6,7] -> 6
Text blocklist management
Following operations are supported to manage your text blocklist:
Create or modify a blocklist
List all blocklists
Get a blocklist by blocklistName
Add blocklistItems to a blocklist
Remove blocklistItems from a blocklist
List all blocklistItems in a blocklist by blocklistName
Get a blocklistItem in a blocklist by blocklistItemId and blocklistName
Delete a blocklist and all of its blocklistItems
You can set the blocklists you want to use when analyze text, then you can get blocklist match result from returned response.
Thread safety
We guarantee that all client instance methods are thread-safe and independent of each other (guideline). This ensures that the recommendation of reusing client instances is always safe, even across threads.
// After you edit your blocklist, it usually takes effect in 5 minutes, please wait some time before analyzing with blocklist after editing.
var request = new AnalyzeTextOptions("I h*te you and I want to k*ll you");
request.BlocklistNames.Add(blocklistName);
request.HaltOnBlocklistHit = true;
Response<AnalyzeTextResult> response;
try
{
response = contentSafetyClient.AnalyzeText(request);
}
catch (RequestFailedException ex)
{
Console.WriteLine("Analyze text failed.\nStatus code: {0}, Error code: {1}, Error message: {2}", ex.Status, ex.ErrorCode, ex.Message);
throw;
}
if (response.Value.BlocklistsMatch != null)
{
Console.WriteLine("\nBlocklist match result:");
foreach (var matchResult in response.Value.BlocklistsMatch)
{
Console.WriteLine("BlocklistName: {0}, BlocklistItemId: {1}, BlocklistText: {2}, ", matchResult.BlocklistName, matchResult.BlocklistItemId, matchResult.BlocklistItemText);
}
}
var getBlockItemId = addedBlockItems.Value.BlocklistItems[0].BlocklistItemId;
var getBlockItem = blocklistClient.GetTextBlocklistItem(blocklistName, getBlockItemId);
Console.WriteLine("\nGet BlockItem:");
Console.WriteLine("BlockItemId: {0}, Text: {1}, Description: {2}", getBlockItem.Value.BlocklistItemId, getBlockItem.Value.Text, getBlockItem.Value.Description);
Remove blockItems
var removeBlockItemId = addedBlockItems.Value.BlocklistItems[0].BlocklistItemId;
var removeBlockItemIds = new List<string> { removeBlockItemId };
var removeResult = blocklistClient.RemoveBlocklistItems(blocklistName, new RemoveTextBlocklistItemsOptions(removeBlockItemIds));
if (removeResult != null && removeResult.Status == 204)
{
Console.WriteLine("\nBlockItem removed: {0}.", removeBlockItemId);
}
Delete text blocklist
var deleteResult = blocklistClient.DeleteTextBlocklist(blocklistName);
if (deleteResult != null && deleteResult.Status == 204)
{
Console.WriteLine("\nDeleted blocklist.");
}
Troubleshooting
General
When you interact with the Azure AI Content Safety client library using the .NET SDK, errors returned by the service will result in a RequestFailedException with the same HTTP status code returned by the REST API request and error code defined by our service. You can parse the RequestFailedException like below:
Analyze text failed.
Status code: 400, Error code: InvalidRequestBody, Error message: The length of given text 1158 exceeds the limit 1000. | Request Id: a04c7c32-ef27-4c23-8b18-07545b24765b, Timestamp: 2023-06-01T16:43:52Z.
Error codes are defined as below:
Error Code
Possible reasons
Suggestions
InvalidRequestBody
One or more fields in the request body do not match the API definition.
1. Check the API version you specified in the API call.<br>2. Check the corresponding API definition for the API version you selected.
InvalidResourceName
The resource name you specified in the URL does not meet the requirements, like the blocklist name, blocklist term ID, etc.
1. Check the API version you specified in the API call.<br>2. Check whether the given name has invalid characters according to the API definition.
ResourceNotFound
The resource you specified in the URL may not exist, like the blocklist name.
1. Check the API version you specified in the API call.<br>2. Double check the existence of the resource specified in the URL.
InternalError
Some unexpected situations on the server side have been triggered.
1. You may want to retry a few times after a small period and see it the issue happens again.<br>2. Contact Azure Support if this issue persists.
ServerBusy
The server side cannot process the request temporarily.
1. You may want to retry a few times after a small period and see it the issue happens again.<br>2.Contact Azure Support if this issue persists.
TooManyRequests
The current RPS has exceeded the quota for your current SKU.
1. Check the pricing table to understand the RPS quota.<br>2.Contact Azure Support if you need more QPS.
Setting up console logging
The simplest way to see the logs is to enable the console logging.
To create an Azure SDK log listener that outputs messages to console use the AzureEventSourceListener.CreateConsoleLogger method.
// Setup a listener to monitor logged events.
using AzureEventSourceListener listener = AzureEventSourceListener.CreateConsoleLogger();
For more extensive documentation on Azure Content Safety, see the Azure AI Content Safety on docs.microsoft.com.
Contributing
This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit cla.microsoft.com.
When you submit a pull request, a CLA-bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., label, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.