Merge pull request #281013 from MicrosoftDocs/main

7/16/2024 PM Publish

Author: Taojunshen

articles/active-directory-b2c/claimsschema.md

@@ -395,7 +395,7 @@ The **Readonly** user input type is used to provide a readonly field to display
 
 #### Paragraph
 
-The **Paragraph** user input type is used to provide a field that shows text only in a paragraph tag.  For example, <p>text</p>. A **Paragraph** user input type `OutputClaim` of self-asserted technical profile, must set the `Required` attribute `false` (default).
+The **Paragraph** user input type is used to provide a field that shows text only in a paragraph tag.  For example, <p>text</p>. A **Paragraph** user input type `OutputClaim` of self-asserted technical profile, must set the `Required` attribute `false` (default). This user input type is only supported in self-asserted page layouts. Unified sign-in and sign-up pages (unifiedssp) might not display this properly.
 
 ![Using claim type with paragraph](./media/claimsschema/paragraph.png)
 

articles/ai-services/openai/concepts/default-safety-policies.md

@@ -0,0 +1,68 @@
+---
+title: Azure OpenAI default content safety policies
+titleSuffix: Azure OpenAI
+description: Learn about the default content safety policies that Azure OpenAI uses to flag content.
+author: PatrickFarley
+ms.author: pafarley
+ms.service: azure-ai-openai
+ms.topic: conceptual 
+ms.date: 07/15/2024
+manager: nitinme
+---
+
+# Default content safety policies
+
+
+Azure OpenAI Service includes default safety applied to all models, excluding Azure OpenAI Whisper. These configurations provide you with a responsible experience by default, including [content filtering models](/azure/ai-services/openai/concepts/content-filter?tabs=warning%2Cpython-new), blocklists, prompt transformation, [content credentials](/azure/ai-services/openai/concepts/content-credentials), and others.
+
+Default safety aims to mitigate risks such as hate and fairness, sexual, violence, self-harm, protected material content and user prompt injection attacks. To learn more about content filtering, visit our documentation describing categories and severity levels [here](/azure/ai-services/openai/concepts/content-filter?tabs=warning%2Cpython-new).
+
+All safety is configurable. To learn more about configurability, visit our documentation on [configuring content filtering](/azure/ai-services/openai/how-to/content-filters).
+
+## Text models: GPT-4, GPT-3.5
+
+Text models in the Azure OpenAI Service can take in and generate both text and code. These models leverage Azure’s text content filtering models to detect and prevent harmful content. This system works on both prompt and completion. 
+
+| Risk Category                             | Prompt/Completion      | Severity Threshold |
+|-------------------------------------------|------------------------|---------------------|
+| Hate and Fairness                         | Prompts and Completions| Medium              |
+| Violence                                  | Prompts and Completions| Medium              |
+| Sexual                                    | Prompts and Completions| Medium              |
+| Self-Harm                                 | Prompts and Completions| Medium              |
+| User prompt injection attack (Jailbreak)  | Prompts                | N/A                 |
+| Protected Material – Text                 | Completions            | N/A                 |
+| Protected Material – Code                 | Completions            | N/A                 |
+
+
+
+## Vision models: GPT-4o, GPT-4 Turbo, DALL-E 3, DALL-E 2
+
+### GPT-4o and GPT-4 Turbo
+
+| Risk Category                                        | Prompt/Completion      | Severity Threshold |
+|------------------------------------------------------|------------------------|---------------------|
+| Hate and Fairness                                    | Prompts and Completions| Medium              |
+| Violence                                             | Prompts and Completions| Medium              |
+| Sexual                                               | Prompts and Completions| Medium              |
+| Self-Harm                                            | Prompts and Completions| Medium              |
+| Identification of Individuals and Inference of Sensitive Attributes | Prompts                | N/A                 |
+| User prompt injection attack (Jailbreak)             | Prompts                | N/A                 |
+
+### DALL-E 3 and DALL-E 2
+
+
+| Risk Category                                     | Prompt/Completion      | Severity Threshold |
+|---------------------------------------------------|------------------------|---------------------|
+| Hate and Fairness                                 | Prompts and Completions| Low                 |
+| Violence                                          | Prompts and Completions| Low                 |
+| Sexual                                            | Prompts and Completions| Low                 |
+| Self-Harm                                         | Prompts and Completions| Low                 |
+| Content Credentials                               | Completions            | N/A                 |
+| Deceptive Generation of Political Candidates      | Prompts                | N/A                 |
+| Depictions of Public Figures                      | Prompts                | N/A                 |
+| User prompt injection attack (Jailbreak)          | Prompts                | N/A                 |
+| Protected Material – Art and Studio Characters    | Prompts                | N/A                 |
+| Profanity                                         | Prompts                | N/A                 |
+
+
+In addition to the above safety configurations, Azure OpenAI DALL-E also comes with [prompt transformation](./prompt-transformation.md) by default. This transformation occurs on all prompts to enhance the safety of your original prompt, specifically in the risk categories of diversity, deceptive generation of political candidates, depictions of public figures, protected material, and others. 

articles/ai-services/openai/concepts/prompt-transformation.md

@@ -0,0 +1,81 @@
+---
+title: Azure OpenAI prompt transformation concepts
+titleSuffix: Azure OpenAI
+description: Learn about the prompt transformation feature in Azure OpenAI DALL-E 3, how it works, and why it's necessary.
+author: PatrickFarley
+ms.author: pafarley
+ms.service: azure-ai-openai
+ms.topic: conceptual 
+ms.date: 07/16/2024
+manager: nitinme
+---
+
+# What is prompt transformation?
+
+Prompt transformation is a process in DALL-E 3 image generation that applies a safety and quality system message to your original prompt using a large language model (LLM) call before being sent to the model for image generation. This system message enriches your original prompt with the goal of generating more diverse and higher-quality images, while maintaining intent.
+
+Prompt transformation is applied to all Azure OpenAI DALL-E 3 requests by default. There may be scenarios in which your use case requires a lower level of enrichment. To generate images that use prompts that more closely resemble your original prompt, append this text to your prompt: `I NEED to test how the tool works with extremely simple prompts. DO NOT add any detail, just use it AS-IS:`. This ensures there is minimal prompt transformation. Evaluating your system behavior with and without this prompt helps you better understand the impact and value of prompt transformation. 
+
+After prompt transformation is applied to the original prompt, content filtering is applied as a secondary step before image generation; for more information, see [Content filtering](./content-filter.md).
+
+> [!TIP]
+> Learn more about image generation prompting in OpenAI's [DALL·E documentation](https://platform.openai.com/docs/guides/images/language-specific-tips).
+
+## Prompt transformation example
+
+
+| **Example text prompt** | **Example generated image without prompt transformation** | **Example generated image with prompt transformation** |
+|---|---|---|
+|"Watercolor painting of the Seattle skyline" | ![Watercolor painting of the Seattle skyline (simple).](../media/how-to/generated-seattle.png) | ![Watercolor painting of the Seattle skyline, with more detail and structure.](../media/how-to/generated-seattle-prompt-transformed.png) |
+
+
+## Why is prompt transformation needed?
+
+Prompt transformation is essential for responsible and high-quality generations. Not only does prompt transformation improve the safety of your generated image, but it also enriches your prompt in a more descriptive manner, leading to higher quality and descriptive imagery.
+
+Default prompt transformation in Azure OpenAI DALL-E 3 contains safety enhancements that steer the model away from generating images of Copyright Studio characters and artwork, public figures, and other harmful content such as sexual, hate and unfairness, violence, and self-harm content.
+
+## How do I use prompt transformation?
+
+Prompt transformation is applied by default to all Azure OpenAI DALL-E 3 requests. No extra setup is required to benefit from prompt transformation enhancements.
+
+Like image generation, prompt transformation is non-deterministic due to the nature of large language models. A single original prompt may lead to many image variants.
+
+
+## View prompt transformations
+
+Your revised or transformed prompt is visible in the API response object as shown here, in the `revised_prompt` field. 
+
+
+```json
+Input Content:
+{
+    "prompt": "Watercolor painting of the Seattle skyline",
+    "n": 1,
+    "size": "1024x1024"
+}
+ 
+Output Content:
+{
+  "created": 1720557218,
+  "data": [
+    {
+      "content_filter_results": {
+        ...
+      },
+      "prompt_filter_results": {
+        ...
+      },
+      "revised_prompt": "A soft and vivid watercolor painting capturing the scenic beauty of the Seattle skyline. The painting illustrates a setting sun casting warm hues over the sprawling cityscape, with the Space Needle prominently standing tall against the sky. Imagine the scattered high-rise buildings, a soothing blend of the lush green of the parks with the winding blue water of the Puget Sound, and the snow-covered peak of Mount Rainier in the distance. A play of light and shadow adds depth and dynamism to this multihued urban panorama."
+    }
+}
+```
+
+> [!NOTE]
+> Azure OpenAI Service does not offer configurability for prompt transformation at this time. To bypass prompt transformation, prepend the following to any request: `I NEED to test how the tool works with extremely simple prompts. DO NOT add any detail, just use it AS-IS:`.
+> 
+> While this addition will encourage the revised prompt to be more representative of your original prompt, the system may alter specific details.
+
+## Next steps
+
+* [DALL-E quickstart](/azure/ai-services/openai/dall-e-quickstart)

articles/ai-services/openai/how-to/dall-e.md

@@ -255,30 +255,10 @@ It's also possible that the generated image itself is filtered. In this case, th
 
 Your image prompts should describe the content you want to see in the image, as well as the visual style of image. 
 
-> [!TIP]
-> For a thorough look at how you can tweak your text prompts to generate different kinds of images, see the [Dallery DALL-E 2 prompt book](https://dallery.gallery/wp-content/uploads/2022/07/The-DALL%C2%B7E-2-prompt-book-v1.02.pdf).
-
-#### [DALL-E 3](#tab/dalle3)
-
 When writing prompts, consider that the image generation APIs come with a content moderation filter. If the service recognizes your prompt as harmful content, it doesn't generate an image. For more information, see [Content filtering](../concepts/content-filter.md).
 
-### Prompt transformation
-
-DALL-E 3 includes built-in prompt rewriting to enhance images, reduce bias, and increase natural variation of images.
-
-| **Example text prompt** | **Example generated image without prompt transformation** | **Example generated image with prompt transformation** |
-|---|---|---|
-|"Watercolor painting of the Seattle skyline" | ![Watercolor painting of the Seattle skyline (simple).](../media/how-to/generated-seattle.png) | ![Watercolor painting of the Seattle skyline, with more detail and structure.](../media/how-to/generated-seattle-prompt-transformed.png) |
-
-The updated prompt is visible in the `revised_prompt` field of the data response object.
-
-While it is not currently possible to disable this feature, you can use special prompting to get outputs closer to your original prompt by adding the following to it: `I NEED to test how the tool works with extremely simple prompts. DO NOT add any detail, just use it AS-IS:`.
-
-#### [DALL-E 2 (preview)](#tab/dalle2)
-
-When writing prompts, consider that the image generation APIs come with a content moderation filter. If the service recognizes your prompt as harmful content, it doesn't generate an image. For more information, see [Content filtering](../concepts/content-filter.md).
-
----
+> [!TIP]
+> For a thorough look at how you can tweak your text prompts to generate different kinds of images, see the [Image prompt engineering guide](/azure/ai-services/openai/concepts/gpt-4-v-prompt-engineering).
 
 
 ## Specify API options

articles/ai-services/openai/toc.yml

@@ -50,6 +50,10 @@ items:
       href: ./concepts/abuse-monitoring.md
     - name: Content filtering
       href: ./concepts/content-filter.md
+    - name: Default safety policies
+      href: ./concepts/default-safety-policies.md
+    - name: Prompt transformations
+      href: ./concepts/prompt-transformation.md
     - name: Customizing Large Language Models (LLMs)
       href: ./concepts/customizing-llms.md
     - name: Embeddings

articles/ai-services/openai/whats-new.md

@@ -20,6 +20,17 @@ This article provides a summary of the latest releases and major documentation u
 
 ## July 2024
 
+### New Responsible AI default content filtering policy 
+
+The new default content filtering policy `DefaultV2` delivers the latest safety and security mitigations for the GPT model series (text), including:
+- Prompt Shields for jailbreak attacks on user prompts (filter), 
+- Protected material detection for text (filter) on model completions 
+- Protected material detection for code (annotate) on model completions
+
+While there are no changes to content filters for existing resources and deployments (default or custom content filtering configurations remain unchanged), new resources and GPT deployments will automatically inherit the new content filtering policy `DefaultV2`. Customers have the option to switch between safety defaults and create custom content filtering configurations. 
+
+Refer to our [Default safety policy documentation](./concepts/default-safety-policies.md) for more information.
+
 ### New GA API release
 
 API version `2024-06-01` is the latest GA data plane inference API release. It replaces API version `2024-02-01` and adds support for:
@@ -57,6 +68,7 @@ Refer to our [data plane inference reference documentation](./reference.md) for
 
 For information on global standard quota, consult the [quota and limits page](./quotas-limits.md).
 
+
 ## June 2024
 
 ### Retirement date updates

articles/ai-studio/how-to/troubleshoot-secure-connection-project.md

@@ -121,6 +121,7 @@ If you use a proxy, it may prevent communication with a secured project. To test
 * Temporarily disable the proxy setting and see if you can connect.
 * Create a [Proxy auto-config (PAC)](https://wikipedia.org/wiki/Proxy_auto-config) file that allows direct access to the FQDNs listed on the private endpoint. It should also allow direct access to the FQDN for any compute instances.
 * Configure your proxy server to forward DNS requests to Azure DNS.
+* Ensure that the proxy allows connections to AML APIs, such as "*.\<region\>.api.azureml.ms" and "*.instances.azureml.ms"
 
 ## Troubleshoot missing storage connections
 

articles/aks/csi-secrets-store-identity-access.md

@@ -201,7 +201,7 @@ In this security model, you can grant access to your cluster's resources to team
     Alternatively, you can create a new managed identity and assign it to your virtual machine (VM) scale set or to each VM instance in your availability set using the following commands.
 
     ```azurecli-interactive
-    az identity create -resource-group <resource-group> --name <identity-name>
+    az identity create --resource-group <resource-group> --name <identity-name>
     az vmss identity assign --resource-group <resource-group> --name <agent-pool-vmss> --identities <identity-resource-id>
     az vm identity assign --resource-group <resource-group> --name <agent-pool-vm> --identities <identity-resource-id>
 

articles/aks/keda-workload-identity.md

@@ -66,7 +66,7 @@ This article shows you how to securely scale your applications with the Kubernet
         --overwrite-existing
     ```
 
-## Deploy Azure Service Bus
+## Create an Azure Service Bus
 
 1. Create an Azure Service Bus namespace using the [`az servicebus namespace create`][az-servicebus-namespace-create] command. Make sure to replace the placeholder value with your own value.