Docker-containerized ML models execute data discovery and classification locally, maintaining data sovereignty while scanning file repositories, SaaS applications, and inline traffic flows, the authors stated.
“Versa DLP uses advanced transformer models and fine-tuned Large Language Models (LLMs) to detect sensitive information across diverse document types and formats. Unlike traditional pattern matching, VersaONE applies contextual and behavioral analysis using spatial and temporal signals to differentiate legitimate data use from potential breaches,” the authors wrote. “This context-aware approach reduces false positives and dynamically applies predefined, policy-driven protections without disrupting normal business workflows.”
In addition, the new version of VersaONE supports inline CASB that inspects traffic and can enforce security policies in real-time. “AI applications are getting updates pretty quickly, or new applications are coming to life, so customers need to be able to detect them quickly and reliably, according to Narin.
Inline CASB lets customers detect these activities, classify them, and apply the right sort of policies onto them. It’s also used for user activity management purposes, Narin said. “For instance, we can say this user is allowed to post certain types of information onto this tool, but another user may not—we can detect many details around how users are using company information,” Narin said.
Versa Verbo, a zero-trust co-pilot
The company has also enhanced VersaONE’s Model Context Protocol server, which allows LLM-powered assistants such Claude, ChatGPT, Gemini, Microsoft Copilot, or internally developed copilots to query Versa systems by accessing Versa APIs. The idea is to make it easier for NetOps and SecOps teams to quickly get answers to security and network issues—without having to switch between multiple tools or consoles, according to Versa.
The company has now integrated the MCP server with Versa’s AI-powered co-pilot, Verbo, to help customers more quickly identify network problems, facilitate troubleshooting, and recommend actions with conversational access to guidance without switching tools.
