Part ofProtection and Enforcement
Protect Sensitive Data in AI Conversations
Detect protected information, apply provider-aware masking, and block risky prompts across OpenAI and ChatGPT, Anthropic Claude, Amazon Bedrock, Azure OpenAI and Azure AI, and Google Gemini.
Product overview
Employees and applications can send sensitive information to Generative AI services in seconds. DataSunrise checks AI interactions against security policy before protected content reaches its destination.
When a rule finds protected data, DataSunrise can mask the matching value or block the prompt. The resulting masking or security event shows what the policy detected and what action it took.
Information Types define the personal, financial, health, credential, and other protected data that each policy should recognize.
How it works
Use one policy flow to inspect an interaction, find protected data, apply a control, and keep the result for review.
DataSunrise reads the prompt or response fields available through the AI service integration.
Information Types check the content for personal, financial, health, credential, and other protected data.
The policy replaces matching values or stops a prompt before the AI service processes it.
Dynamic Masking Events and Security Events show what the rule found and what action it took.
Generative AI Activity Monitoring can add the surrounding interaction and identity context for investigation.
Generative AI protection capabilities
Masking replaces the sensitive part of an interaction while allowing the rest of the content to continue. Prompt controls stop a matching request before the AI service processes it.
Replace protected values in the prompt or response fields available through the AI service integration.
Stop a request when its content violates a security rule, then record the event for review.
DataSunrise applies the same sensitive-data policy model while working with the fields and traffic format exposed by each AI service.
| AI service | Available controls |
|---|---|
| OpenAI and ChatGPT | Block risky prompts and mask sensitive response content |
| Amazon Bedrock | Block prompts and mask sensitive request or response content |
| Azure OpenAI and Azure AI | Block risky prompts and mask sensitive response content |
| Anthropic Claude | Block risky prompts and mask sensitive Claude fields |
| Google Gemini | Inspect sensitive data, block risky prompts, and mask protected content |
Information Types tell DataSunrise what to look for. Teams can start with built-in definitions, add custom patterns and dictionaries, or tune the attributes that identify data specific to the business.
The same definition can support audit, masking, and security rules, so teams do not have to describe the same protected data again for every policy.
Dynamic Masking Events show when a rule replaced sensitive content. Security Events show prompts stopped by a security rule. Each record gives teams the policy, AI service, available identity context, and action needed to understand what happened.
Generative AI Activity Monitoring adds the surrounding interaction history for investigation.
Activity Monitoring records AI interactions for investigation. Data Masking and Prompt Controls can act on the interaction by replacing sensitive content or stopping a matching prompt.
No. This product protects traffic sent to Generative AI services. AI-Assisted Data Masking uses a locally configured model to generate replacement values for DataSunrise masking workflows.
Yes. A security rule can block a matching prompt and record the event for review.
The available direction depends on the fields exposed by the AI service. The AI services tab above shows the controls available for each integration.
Map the right workflow
