PromptGuard AI DLP

Privacy notice before installation: PromptGuard processes text entered into supported prompt fields locally in the browser. Prompt text and detected values are not sent to Fragenta or to the company workspace. When connected to a workspace, PromptGuard sends only the supported AI hostname, protection action, detected data categories, match count, policy version, time, a user-selected device name, and a pseudonymous device assignment over HTTPS. These metadata are used only for the described protection, device management, and security audit. PromptGuard helps companies prevent sensitive text from being accidentally sent to supported AI services. It checks input directly in the browser and enforces the organization’s policy before transmission. Features in the controlled B2B pilot: • Local detection of IBANs, credit card numbers, API keys, private-key headers, password contexts, and selected personal data • Warning, anonymization, or blocking before transmission • Protection for ChatGPT, Gemini, Microsoft Copilot, and Claude in managed desktop browsers • Centrally managed, versioned company policies • Minimized audit metadata containing category, action, AI service, and time—never prompt text or the detected value • A visible protection status and a strict local fallback if the policy service is unavailable PromptGuard protects supported websites in the managed browser. The current pilot does not cover native applications, direct API calls, images, screenshots, or file uploads. Strict local baseline protection works without an account. A PromptGuard workspace provided by the organization is required for centrally managed policies, device management, and minimized audit metadata. PromptGuard is an independent product and is not affiliated with OpenAI, Google, Anthropic, or Microsoft. All trademarks belong to their respective owners.