Request a demo

AI at work without data leaks

Niqob replaces names, card and phone numbers with codes before a prompt reaches ChatGPT, Claude or DeepSeek, and puts them back in the reply. Employees use AI as usual, the data stays in the company.

Request a demoHow it works

The employee writes

The AI receives

Make a bonus table: [NAME_1] [SURNAME_1], card [CARD_1], phone [PHONE_1]. [NAME_2] [SURNAME_2], PINFL [PINFL_1].

The employee gets the reply

Done, here is the table:

  • First name Alisher
  • Last name Karimov
  • Card 8600 1234 5678 9012
  • Phone +998 90 123 45 67
  • First name Dilnoza
  • Last name Rahimova
  • PINFL 31203901234567
Edit the text: the check runs in your browser and nothing is sent anywhere. The data in the example is made up.

AI has become a work tool and the main channel for data leaks

  • 32%

    of all unauthorized data movement out of companies goes through AI services

    LayerX, 2025

  • 82%

    of pastes into AI chats come from personal accounts the company cannot see

    LayerX, 2025

  • 34.8%

    of the corporate data employees enter into AI is sensitive

    Cyberhaven, 2025

Bans get ignored, and working without control costs more

  • A ban slows work down

    Samsung

    banned AI after engineers uploaded source code to ChatGPT. Citigroup, Goldman Sachs and JPMorgan also restricted it.

    Bloomberg / Insider, 2023

  • Bans are ignored

    57%

    of employees entered sensitive data into public AI, and 29% knew it was against company rules.

    TELUS Digital, 2025

  • No control is expensive

    +$670K

    added to the cost of a breach when shadow AI is involved. 97% of affected companies had no AI access controls.

    IBM, 2025

Employees use any AI. The data never leaves the company.

No bans and no loss of speed, for corporate subscriptions and personal accounts alike.

The AI gets the task, but not the data

  1. 1

    Detect

    Before sending, Niqob checks the prompt for names, cards, PINFL, phones, tax IDs and your company’s own rules.

  2. 2

    Replace

    The data is swapped for codes like [NAME_1], and the prompt gets an instruction on how to work with them.

  3. 3

    Restore

    Codes in the reply are swapped back on the company side. The employee gets a finished result.

Where masking does not fit, Niqob blocks the prompt, and control is rolled out gradually

  • Masking

    Data is replaced with codes, work keeps going.

  • Blocking

    The prompt never reaches the AI. The employee sees why right away and can resend it without the sensitive data.

  • Strict level

    Every prompt is checked on the fly.

  • Gradual level

    Analysis first, then filters. Speed does not drop at the start, and control reaches 100% as filters are tuned.

See the plugin at work in Claude

A recording of the demo plugin in real use: data turns into codes right in the input box, and the reply comes back with the original data.

The video will appear here soon.

Demo plugin for Chrome Demo 0.2.0

Works on claude.ai. Filters and codes stay inside the extension, no server needed.

  1. Unzip the archive into its own folder.
  2. Open chrome://extensions and turn on Developer mode.
  3. Click Load unpacked and pick that folder.
Download the demo (ZIP)

One core, five ways to connect

  • Browser plugin

    AI chats in the browser

  • Device agent

    Desktop apps

  • Apps with AI

    Software with built-in AI

  • API

    Your internal systems

  • Subscriptions through us

    Models without separate subscriptions

Niqob core: check, mask, block

ChatGPT, Claude, DeepSeek, Gemini and other AI

Niqob runs in our cloud or entirely on your servers

Our cloud

  • Quick start with no infrastructure of your own
  • Proxy and processing on our servers and domains

Your server

  • For strict data storage requirements
  • Workers and proxy run on your side, we only check the license and send settings
  • We have no access to your database

Demand for this kind of product is proven, and the niche in Uzbekistan is open

Security leaders are buying this class of product

CompanyBuyerAmount
Prompt SecuritySentinelOne~$250M
LakeraCheck Point~$300M
Protect AIPalo Alto Networks$634.5M
Calcalist, Entrepreneur, SEC 10-K

×30

growth in Russia in 2025 of confidential data sent to public neural networks. No data masking was found among local vendors.

Solar, 2026

Uzbekistan

  • Law

    Criminal Code art. 141.2: liability for unlawful transfer of personal data over the internet. Since March 2026, biometrics and telecom subscriber data must be stored in the country.

  • Market

    AI strategy to 2030: AI products and services up to $1.5B. 35 commercial banks, sector profit up 57% in 2025.

  • Niche

    Classic DLP exists, but no public solutions for AI chats were found.

Anhor, gov.uz, Dunyo, Golden Pages

Niqob matters most where a leak costs the most

  • Banks

    Cards, customer personal data, bank secrecy. Global banks have already restricted AI.

  • Telecom

    By law, subscriber data is stored only in Uzbekistan.

  • Retail

    Customer bases and payment data, many employees across offices and stores.

Built for companies where hundreds or thousands of employees work with AI.

The price is calculated for your company: you pay only for what we protect

What it depends on

  • Number of employees
  • AI services and connection methods
  • Protection level
  • Our cloud or your server

How to get an exact figure

  1. Analysis of the company’s AI traffic
  2. Connecting one work process
  3. An offer based on real data

+$670K

the cost of doing nothing: what shadow AI adds to the cost of a single breach on average. IBM, 2025

We’ll show Niqob on your tasks

Tell us where your employees use AI. We’ll prepare a demo on your scenarios or a solution for the whole company. The first process is connected in under two weeks.

  1. 1

    Prepare

    For your company: processes, AI services, sensitive data.

  2. 2

    Connect

    A group of employees or the whole company, through the browser plugin, agent or API.

  3. 3

    Show

    Which data would have gone to AI and how Niqob covers it.

Request a demo or a solution

What you need

Sources

Most figures come from security vendors’ research, and each one names its author.

  1. LayerX — Enterprise AI & SaaS Data Security Report 2025
  2. Cyberhaven — AI Adoption and Risk Report 2025
  3. TELUS Digital — survey, 2025
  4. IBM — Cost of a Data Breach 2025
  5. Samsung — Bloomberg / Insider, 2023
  6. Prompt Security → SentinelOne — Calcalist, 2025
  7. Lakera → Check Point — Entrepreneur, 2025
  8. Protect AI → Palo Alto Networks — SEC 10-K
  9. Солар — утечки в публичные нейросети, 2025
  10. Anhor — локализация персональных данных, 2026
  11. Anhor — ст. 141.2 УК Узбекистана
  12. gov.uz — Стратегия ИИ до 2030 (ПК-358)
  13. Dunyo — банковский сектор Узбекистана, 2025
  14. Golden Pages — Datagaze