DAC for LLM Observability

config/_default/menus/main.en.yaml

@@ -4713,16 +4713,21 @@ menu:
       parent: llm_obs
       identifier: llm_obs_experiments
       weight: 5
+    - name: Data Security and RBAC
+      url: llm_observability/data_security_and_rbac
+      parent: llm_obs
+      identifier: llm_obs_data_security_and_rbac
+      weight: 6
     - name: Terms and Concepts
       url: llm_observability/terms/
       parent: llm_obs
       identifier: llm_obs_terms
-      weight: 6
+      weight: 7
     - name: Guides
       url: llm_observability/guide/
       parent: llm_obs
       identifier: llm_obs_guide
-      weight: 7
+      weight: 8
     - name: CI Visibility
       url: continuous_integration/
       pre: ci

content/en/account_management/rbac/data_access.md

@@ -43,10 +43,10 @@ Log in as a user assigned the Datadog Admin role, or any user with a role in you
 
 In order to create a Restricted Dataset, identify the data to be restricted with a query.
 
-{{< img src="/account_management/rbac/restricted_dataset.png" alt="Create a Restricted Dataset dialog. Selects data in RUM, APM, Logs, and Metrics matching the tag service:hr. Grants access to a Privileged access team.">}}
+{{< img src="/account_management/rbac/restricted_dataset-2.png" alt="Create a Restricted Dataset dialog. Selects data in RUM, APM, Logs, and Metrics matching the tag service:hr. Grants access to a Privileged access team.">}}
 
 Name Dataset
-: A descriptive name to help users understand what data is contained in the dataset. 
+: A descriptive name to help users understand what data is contained in the dataset.
 
 Select data to be included in this Dataset
 : The boundary definition that describes which data to restrict to a specific set of users. Boundaries are query statements with limitations that allow an access manager to define the scope of sensitive data to be protected. The [supported telemetry types][10] are custom metrics, RUM sessions, APM traces, logs, cloud costs, error tracking issues, and CI Visibility pipelines.
@@ -61,7 +61,7 @@ After completing all the fields to define the dataset, click **Create Restricted
 You may create a maximum of 100 Restricted Datasets. If you need a higher limit, reach out to Support.
 
 ### API
-The Data Access Control API is under development and should be considered unstable. Future versions may be backward incompatible. 
+The Data Access Control API is under development and should be considered unstable. Future versions may be backward incompatible.
 
 Terraform support will be announced after Data Access Control is generally available.
 
@@ -75,6 +75,7 @@ Terraform support will be announced after Data Access Control is generally avail
 - Error Tracking issues
 - Logs
 - RUM sessions
+- LLM Observability
 
 ## Usage constraints
 
@@ -92,6 +93,7 @@ Playlists are collections of Session Replays you can aggregate in a folder-like
 ### Logs
 Data Access Control is separate from the existing [Logs RBAC permissions][11] feature, also known as log restriction queries. To use Data Access Control with Log Management, first request access to Data Access Control. Next, manually migrate your configuration from Log Management permissions to Data Access Control.
 
+
 ## Select tags for access
 
 Each Restricted Dataset can control access to multiple types of data, such as metrics. You are free to use the same or different tags across multiple types of telemetry. Within each telemetry type, you must use a _single_ tag or attribute to define your access strategy.
@@ -101,7 +103,7 @@ If you have too many combinations of tags or attributes to fit within these cons
 ### Supported example
 
 #### Restricted Dataset 1
-- Telemetry Type: RUM  
+- Telemetry Type: RUM
    - Filters: `@application.id:ABCD`
 
 #### Restricted Dataset 2
@@ -112,23 +114,23 @@ If you have too many combinations of tags or attributes to fit within these cons
 
 ### Not supported example
 
-#### Restricted Dataset 1: 
-* Telemetry type: RUM 
+#### Restricted Dataset 1:
+* Telemetry type: RUM
     * Filters: `@application.id:ABCD`
 
 #### Restricted Dataset 2:
-* Telemetry type: RUM 
+* Telemetry type: RUM
     * Filters: `env:prod`
 
 Restricted Dataset 1 uses `@application.id` as the tag for RUM data, so a new Restricted Dataset can't change to a different tag. Instead, consider reconfiguring Restricted Dataset 2 to use `@application.id`, or changing all of your Restricted Datasets with RUM data to use another tag.
 
 ### Not supported example
 
-#### Restricted Dataset 1: 
-* Telemetry type: RUM 
+#### Restricted Dataset 1:
+* Telemetry type: RUM
     * Filters: `@application.id:ABCD`
 
-#### Restricted Dataset 2: 
+#### Restricted Dataset 2:
 * Telemetry type: RUM
     * Filters: `@application.id:IJKL` `env:prod`
 
@@ -150,7 +152,7 @@ For example, if you have a single application that is instrumented with Real Use
     * Telemetry type: RUM
         * Filters: `@application.id:<rum-app-id>`
 * **Grant access:**
-    * Teams or roles of users who can see this RUM data      
+    * Teams or roles of users who can see this RUM data
 
 This configuration example would protect the RUM data from this application, and keep other data from this application available to existing users in your organization.
 

content/en/llm_observability/data_security_and_rbac.md

@@ -0,0 +1,23 @@
+---
+title: Data Security and RBAC
+further_reading:
+  - link: "/account_management/rbac/data_access"
+    tag: "Documentation"
+    text: "Learn more about data access controls"
+---
+{{< whatsnext desc=" ">}}
+  {{< nextlink href="https://datadoghq.com/legal/hipaa-eligible-services">}}<u>HIPAA-Eligible Services</u>: Datadog Legal's list of HIPAA-eligible services{{< /nextlink >}}
+  {{< nextlink href="/llm_observability/evaluations/#sensitive-data-scanner-integration">}}<u>Sensitive Data Scanning for LLM Observability</u>: Redact sensitive information in your LLM application{{< /nextlink >}}
+{{< /whatsnext >}}
+
+## Data Access Control
+
+LLM Observability allows you to restrict access to potentially sensitive data associated with your ML applications to only certain teams and roles in your organization. This is particularly important when your LLM applications process sensitive information such as personal data, proprietary business information, or confidential user interactions.
+
+Access controls in LLM Observability are built on Datadog's [Data Access Control][1] feature, which enables enables you to regulate access to data deemed sensitive. You can use the `ml_app` tag to identify and restrict access to specific LLM applications within your organization.
+
+## Further reading
+
+{{< partial name="whats-next/whats-next.html" >}}
+
+[1]: /account_management/rbac/data_access