Skip to content

Stockertim Dictionary · Artificial intelligence · Cybersecurity

Data loss prevention

Also called DLP, data leakage prevention

In short

Data loss prevention is the set of tools and rules that detect and block sensitive data leaving an organisation through email, uploads, prompts or other channels.

  • Intermediate
  • Technology
  • Checked

In detail

Data loss prevention, usually called DLP, watches for sensitive information on the move and stops or flags it. It looks for patterns such as card numbers, national insurance numbers, bank details or files marked confidential, and checks them against rules about where that kind of data may go. A match can trigger a warning to the user, a block, or an alert to the security team.

DLP tools sit at several points: on staff devices, in email systems, at the edge of the network and inside cloud services. Many organisations now extend them to AI tools, checking what is pasted into prompts or uploaded to an assistant and blocking unapproved services. Classification labels, such as Public, Internal and Confidential, make the rules far more accurate.

DLP matters because most data leaks with AI are accidental: someone pastes a client spreadsheet into an unapproved tool to save time. A well-tuned system catches many of these without slowing everyday work. A badly tuned one floods the team with false alarms, so staff learn to ignore warnings, which is why tuning and review are ongoing tasks.

DLP is one layer, not a complete answer. It works best alongside clear rules about approved tools, training, and making the safe option the easy one.

Worked example

Harbour Foods reviews one month of DLP alerts linked to AI tools.

  1. Alerts raised: 240, of which 18 were real attempts to send sensitive data
  2. Share of alerts that were real: 18 / 240 = 7.5%
  3. Of the 18, 15 were client lists pasted into an unapproved assistant: 15 / 18 = 83.3%
  4. After tuning two rules, false alarms fall from 222 to 120: (222 − 120) / 222 = 45.9% fewer

The review shows the main leak route, and tuning cuts false alarms by almost half so real alerts are easier to spot.

In practice

Security teams apply classification labels, set DLP rules for email, uploads and AI tools, review alerts weekly, tune rules that cause noise and feed real cases into training.

Watch out

  • DLP does not see everything, such as photos of a screen or data typed from memory.
  • A high number of alerts is not success; what matters is how many real leaks are caught and how few false alarms remain.

Learn it properly

Official sources

Facts last checked on 3 November 2026.

Education, not advice: the Stockertim Dictionary explains what terms mean. It never recommends buying or selling anything.

English