Anzenna launches AI DLP for prompts and agent activity

Anzenna on July 30, 2026 launched AI DLP to detect and stop sensitive data from reaching sanctioned and shadow AI tools. The product is designed to reduce false positives, score behavior against identity context, and give security teams real-time options to block, quarantine, revoke access, or coach users.

Why it matters: - AI tools have created new data-loss paths that legacy DLP was not built to inspect. - Anzenna is targeting prompts, file uploads, OAuth grants, MCP servers and coding agents, where sensitive data can leave the enterprise through approved and shadow AI systems. - The company says the approach can cut alert noise and help security teams focus on real risk instead of benign activity.

What happened: - Anzenna announced AI DLP on July 30, 2026. - The capability detects and stops sensitive data before it reaches AI tools. - The product is available now. - Anzenna says AI DLP uses endpoint and identity infrastructure customers already have. - The company directs interested customers to learn more or request a demonstration.

The details: - AI DLP correlates identity and data signals across the enterprise to find sensitive content headed to AI tools, including sanctioned and shadow apps. - The system covers prompts, file uploads, MCP activity, OAuth grants and agent activity. - Each movement is scored against the employee’s behavioral baseline, role, peer group and the sensitivity of the data. - The scoring replaces static rule-based decisions. - Violations are turned into prioritized case files with the identity, data type, destination tool and reasoning behind the verdict. - Anzenna says customers see roughly 90% fewer alerts than with existing DLP tooling. - Security teams can block the transfer, quarantine the data, revoke access or coach the user. - Every action keeps a full audit trail.

Between the lines: - Legacy DLP products were built around files moving through email and endpoints, not AI prompts and agent workflows. - The pitch is not just detection. It is contextual decision-making tied to who moved the data and whether the behavior fits that person. - That framing is meant to separate normal employee use from risky behavior, such as using a personal AI account to move faster or a departing engineer pulling data they have not touched before. - A quoted customer example said legacy DLP produces more than 1,000 alerts a day and misses AI coverage, while Anzenna reduces alert volume by 90%.

What's next: - Security teams can use the product to intervene in real time through existing endpoint and identity systems. - Anzenna says the goal is to produce evidence-backed investigation reports in under two minutes. - The company will likely lean on AI governance and insider-risk use cases as adoption expands across browsers, IDEs, MCP servers and OAuth-authorized AI apps.

The bottom line: - Anzenna is trying to make DLP work for the AI era by combining data inspection with identity-aware behavior scoring.