The core of the Cyphlens technology is in-document encryption that functions at a granular level. When a user tasks an AI with identifying a renewal date in a legal agreement, the software can mask the negotiated price while leaving the remaining text readable for the model. Crucially, these redactions occur in-place, meaning the file remains a standard Word or PDF format rather than an encrypted container. This allows a single document to reveal different information sets to different users or AI agents depending on their authorization.

Beyond document parsing, the company offers a capability called CypherFile, which manages protection for files up to one terabyte. The system operates on a zero-knowledge architecture where content never reaches Cyphlens servers; encryption and decryption occur locally on the sender's device. This ensures that even after a file is shared and travels across various channels, the sender retains the ability to revoke access. If a document is recalled, the content becomes permanently inaccessible because each attempt to open the file requires a real-time verification check.

CEO Rocky Motwani notes that the traditional model of security—assuming the person opening a file is the only one reading it—has collapsed in the age of generative AI. By providing a control layer that governs exactly what a human or a machine can see, the firm aims to allow organizations to integrate AI into document-heavy workflows without sacrificing governance. Salesforce Ventures partner Rob Keith highlighted this shift, noting that effective security now centers on granular visibility, ensuring data remains protected even after it leaves a company’s internal perimeter.