Praeferre/Platform/AI Data Leak Protection
AI Data Leak Protection

Let your teams use AI. Keep your data out of it.

Praeferre sits between your people and every AI provider they reach for — ChatGPT, Claude, Gemini and beyond — redacting, obfuscating or blocking sensitive data on the way out, and giving you full visibility into how AI is actually being used across your workspace.

Protecting data shared with: CHATGPT CLAUDE GEMINI + ANY AI TOOL IN YOUR WORKSPACE
Why it matters

Your people are already pasting into AI. The question is what goes with it

Shadow AI isn't a hypothetical risk — it's the default state of most workplaces today. Employees aren't being reckless; they're trying to work faster, using tools that feel no different from search. But every prompt is a potential export of data you don't control once it leaves.

Case study · Samsung Electronics, 2023

Three leaks in twenty days

Within weeks of engineers gaining access to ChatGPT, Samsung suffered three separate exposures of confidential material — with no malicious intent behind any of them.

Day 1An engineer pastes proprietary semiconductor source code into ChatGPT, asking it to fix a bug.
Day 9A second engineer submits equipment-defect detection code, seeking optimisation suggestions.
Day 20A third employee transcribes an internal meeting and feeds the full transcript in for meeting notes.

Samsung's response was an emergency company-wide ban — the same blunt instrument most organisations reach for once a leak is already public, because a policy alone couldn't stop it in time.

€15M
OpenAI · Italy's Garante, 2024

Fined for processing personal data to train ChatGPT without an adequate legal basis, and for failing to report a March 2023 breach. The decision was later annulled on jurisdictional grounds in 2026 — but the underlying findings on legal basis and transparency were never overturned on their merits.

$670K
Added breach cost · IBM, 2025

Average additional cost when shadow AI was a factor in a breach — on top of the $4.44M global average, and with customer PII exposed at a higher rate than the norm.

Sources: Dark Reading, on the Samsung ChatGPT incidents · Euronews, on the Garante's OpenAI fine · IBM Cost of a Data Breach Report, 2025.

66%

Of office professionals have used AI tools at work despite believing it wasn't permitted under company policy — and more than a third entered customer data into public models.

97%

Of AI-related breaches involved systems that lacked proper access controls — and most affected organisations had no governance policy for AI use at all.

4x

Year-on-year rise in shadow AI detections in enterprise environments, now the third most common non-malicious insider risk observed.

Sources: PagerDuty Shadow AI Survey, 2026 · IBM Cost of a Data Breach Report, 2025 · Verizon 2026 Data Breach Investigations Report, via Tech Times.

How it works

Detect. Decide. Teach.

Praeferre doesn't just block AI — it makes each interaction a safe one, and a chance for the employee to learn why.

STEP 1 · DETECT

Every prompt is inspected before it sends

The browser extension and desktop agent read prompts, uploads and pastes in real time — across web AI tools and the desktop apps that talk to them — spotting PII, credentials, financial data and contract language as it's typed.

Browser extensionDesktop agentReal time
STEP 2 · DECIDE

Redact, obfuscate, block or allow

Every category of sensitive data has its own policy — set by you. Some data gets quietly redacted, some gets synthetically obfuscated, some is blocked outright, and low-risk content simply passes through.

Per data typePer teamPer AI tool
STEP 3 · TEACH

Every interception is a coaching moment

When something is redacted or blocked, the employee sees why, in plain language — turning enforcement into on-the-spot upskilling instead of a silent wall, with the full history feeding your reporting.

In-context guidanceManager reportingTrend tracking
Set your own rules

Redact, obfuscate or block — your call, by data type

No two organisations treat data the same way, so Praeferre doesn't force one policy on everyone. Set exactly how each category of sensitive information is handled, and adjust it per team, department or AI provider as your risk appetite changes.

  • Redact — strip the sensitive value, let the rest of the prompt through
  • Obfuscate — swap in realistic synthetic data so the prompt still works
  • Block — stop the submission entirely, with a clear reason shown
Configure your policy
Data Handling Preferences
Customer PII (names, emails, phone)
REDACT
Financial identifiers (IBAN, card numbers)
BLOCK
Contract & legal text
BLOCK
Source code snippets
OBFUSCATE
Internal project names
OBFUSCATE
General written content
ALLOW
Full visibility

See how AI is actually being used, tool by tool

You can't govern what you can't see. Praeferre gives you a live view across your workspace — which AI providers are being used, by whom, how often, and what got caught along the way — so "shadow AI" stops being a guess and becomes a dashboard.

  • Usage broken down by team, individual and AI provider
  • Full audit trail of every redaction, obfuscation and block
  • Feeds straight into your GRC evidence base — no separate export
See a live dashboard
Workspace AI Activity
Finance teamChatGPTIBAN redacted from a supplier email draft2m ago
Legal teamClaudeContract clause blocked — outside workspace policy14m ago
EngineeringGeminiSource code snippet obfuscated before send41m ago
MarketingChatGPTDraft copy allowed — no sensitive data found1h ago
HRClaudeEmployee record redacted from a summary request2h ago
Document & contract intelligence

Contracts don't leave the workspace — even by accident

Beyond the prompt box, Praeferre's models scan the documents your teams work with — contracts, agreements, board papers — and understand what they are, not just what they contain. A clause pasted from a live vendor contract is treated differently to a paragraph from a public brochure, because the model recognises the document type itself.

  • Automatic classification of contracts, NDAs and legal documents
  • Blocks contract content from leaving the workspace via any AI tool
  • Works on pasted text, uploaded files and copied excerpts alike
See document scanning in action
Document Classification
vendor_msa_2026.docx pasted excerpt product_brochure.pdf pasted excerpt CLASSIFIED Legal · Contract CLASSIFIED Marketing · Public BLOCKED at source ALLOWED low risk
Approval workflows & upskilling

Turn every block into a learning moment

A blanket ban teaches nothing and pushes people toward personal accounts you can't see at all. Praeferre routes edge cases to an approver, explains every decision in plain language, and rolls the pattern into reporting your managers can actually use — so your organisation gets measurably safer with AI over time, not just more restricted.

Approval workflow

Escalate, don't just block

Borderline requests route to a named approver instead of a dead end — so legitimate work isn't blocked by an over-cautious policy, and every exception is logged.

In-context coaching

Explain the "why," every time

When something is redacted or blocked, the employee sees exactly which policy triggered it and why — building understanding instead of workaround instincts.

Manager reporting

See where training would help most

Team-level trends surface which departments or individuals trigger the most interceptions — turning the platform into a targeted upskilling programme, not just a filter.

Start your compliance journey

See what's leaving your workspace right now

A short discovery call is enough to show you what a live scan of your AI usage would surface in week one — no ban required.