CloakWellAI

CloakwellAI is an enterprise-grade DLP (Data Loss Prevention) platform that solves a critical security challenge: employees inadvertently leaking corporate secrets and PII to public cloud AI services like ChatGPT, Claude, and Gemini. By acting as an intelligent security layer between users and cloud AI, Cloakwell runs a sophisticated two-tier detection pipeline. The first tier employs fast client-side pattern matching to catch obvious PII: credit card numbers, API keys, passwords, and credentials. The second tier leverages Gemma 2 (2B) running on dedicated AMD Instinct GPUs to perform advanced contextual classification and identify sophisticated leaks, including internal project codenames, infrastructure details, employee directories, and cumulative information exposure across entire conversations. When sensitive data is detected, Cloakwell doesn't simply block requests—it intelligently redacts sensitive information. Using advanced named entity recognition, the platform surgically removes PII and replaces it with semantic placeholders like [REDACTED_CREDENTIAL], and [REDACTED_PROJECT_NAME], preserving prompt usability and meaning while guaranteeing zero data egress. The cleaned, redacted query is then safely routed to public cloud AI services for maximum performance and capability, enabling employees to work efficiently without compromising security. The platform provides comprehensive coverage across multiple interfaces: browser extensions protect web UIs (ChatGPT, Claude), while a TLS-intercepting proxy secures CLI and agentic developer tools like Claude Code. A real-time web-based dashboard enables monitoring of redactions, blocks, and complete audit logs of all data classifications and routing decisions. Built entirely on open-source technology and AMD Instinct hardware, Cloakwell enables enterprises to adopt AI confidently without sacrificing DLP compliance, data governance, or security posture. This is the future of secure AI adoption.