Barracuda AI Data Security Moves DLP Into Prompts

Barracuda AI Data Security inspects prompts and uploads across more than 1,300 GenAI tools, with free visibility and paid enforcement tiers.

Key takeaways

The product inspects prompts and uploaded files before they reach GenAI services.
Barracuda says coverage spans more than 1,300 GenAI tools.
Visibility starts free; enforcement and broader workflows are paid tiers.

What happened

Barracuda has launched AI Data Security, a control layer designed to inspect employee prompts and file uploads before sensitive information reaches generative-AI services. The company says the product can discover unsanctioned use, apply user or group policies, and block, warn or redact content such as credentials, source code, payment data and health records. Its paid coverage is described as spanning more than 1,300 GenAI tools, including ChatGPT, Claude, Gemini and Microsoft Copilot.

Why Barracuda AI Data Security matters

Traditional data-loss prevention often focuses on email, cloud storage and managed applications. Generative AI creates another outbound path: a useful prompt can also contain customer records, proprietary code or credentials. Barracuda is moving that inspection point closer to the prompt. The business argument is simple—allow approved AI use while making data handling observable and enforceable. The harder question is whether the controls can distinguish risky context from harmless text without overwhelming employees with warnings.

The packaging starts with visibility

The Foundation tier is free and is positioned as a way to discover shadow-AI usage and data-security risk. Premium adds policy enforcement across AI tools, while Premium Plus extends real-time prompt inspection and DLP to additional tools and workflows. Barracuda does not publish list prices for the paid tiers. Existing SecureEdge Access customers can extend familiar web inspection and identity-aware policies without a separate deployment project, according to the product page.

Claims need operational proof

This is a product launch, not proof that every leak will be stopped. Coverage breadth does not automatically establish detection quality, latency or employee acceptance. Buyers should test how the service handles code snippets, multilingual prompts, encrypted uploads, personal accounts and sanctioned tools with different retention settings. They should also ask where inspection data is processed, how long logs are retained and which administrators can read prompt content. Those details determine whether a security control creates a new privacy problem.

What enterprises should measure

A useful pilot should record the number of discovered tools, high-confidence sensitive-data events, false positives, override rates and time added to routine work. Security teams should pair those metrics with an approved-tool policy and an incident path for genuinely exposed data. The launch is notable because it turns AI governance from a policy document into an inline control. Its value will depend on precise detection, explainable decisions and a rollout that does not simply push staff toward unmanaged devices.

Product claims become useful only after workload testing and fit analysis.

Facts table

Launch date	22 September 2026
Named coverage	More than 1,300 GenAI tools
Entry tier	Foundation, free
Paid tiers	Premium and Premium Plus
Policy actions	Block, warn or redact