Summary
iboss has introduced AI DLP, a data loss prevention offering designed to detect, flag, and block sensitive data headed for AI platforms, cloud applications, or other cloud services. The offering addresses sensitive data that legacy approaches fail to catch.
Sensitive data leaves through files shared to cloud drives, forms posted to web applications, and source code pasted into AI tools. Much of it has no fixed format, so rules that search for account numbers or ID strings let it pass. Data labels create a second gap, because files that are mislabeled or never labeled fall outside policy and leak.
AI DLP applies AI/ML classification and deep content inspection to read what data contains as it moves. File analysis and OCR read documents and text inside images, and classification identifies sensitive data whether it is labeled, mislabeled, or unlabeled. Where Microsoft Information Protection labels exist, the product reads and enforces them inline.
Policies set by user, group, and department detect, flag, or block source code, customer data, PII, and intellectual property in real time, with a full audit trail. These policies cover AI platforms and tools, so teams can keep sensitive data out of AI without banning them. They also apply when AI agents move data, with control over permitted AI services, tools, and APIs and retained agent activity.
AI DLP runs on the iboss containerized architecture, processing each customer's traffic in dedicated containers with isolated SSL keys and dedicated IPs, and customer data is never co-mingled. It works alongside the iboss Secure Web Gateway, CASB, GenAI Protection, and Zero Trust Network Access, with one console, one policy engine, and one data lake. Risk scoring and one-click GDPR, HIPAA, PCI-DSS, and SOX reports give auditors evidence of what was found and stopped.
“Data labels are a challenge for DLP. Files get mislabeled, and data that is never labeled leaks,” said Paul Martini, CEO and Co-Founder of iboss. A live demonstration webinar is scheduled for November 5, 2026, and a private preview is set for December 2026.
Our summary. The original keeps its own byline.