Data Loss Prevention - Define custom topics for AI prompt protection

You can now define custom topics for AI prompt protection. Predefined AI prompt topics cover common content and intent categories such as PII, source code, and jailbreak attempts. Custom topics let you detect unique or proprietary concepts that are not included in predefined categories. You describe a custom topic in natural language, and Cloudflare DLP detects whether a prompt matches that topic based on context rather than specific keywords. For example, a topic that describes confidential merger discussions matches a prompt that paraphrases the deal, even when the prompt never uses the word merger or names the companies involved. To detect literal values such as internal codenames or product identifiers, use a custom wordlist or pattern entry instead. Custom topics run through the same application granular controls path as predefined AI prompt topics. Custom topics are available for ChatGPT, Google Gemini, Perplexity, and Claude. Create a custom AI prompt topic In the Cloudflare dashboard , go to Zero Trust > Data loss prevention > Detection entries . Select AI prompt topics , then select Custom Prompt Topic . Describe the topic in natural language. Be specific about the concept you want to detect. For example, describe unreleased product roadmap details or confidential customer contract terms. Add this detection entry to an existing DLP profile, or create a new DLP profile . Use the profile in a Gateway HTTP policy to log or block prompts that match the topic. Note Write the description as a concept to classify, not a list of keywords. For example, describe "internal financial forecasts and unreleased revenue figures" rather than listing specific document names. For more information, refer to AI prompt topics .