Safe use of AI at work

task2mask replaces names, numbers and passwords with codes before anything leaves the company. Employees keep using ChatGPT, Claude and Gemini as before, and company data stays inside.

The problem

Employees paste company data into AI tools

Almost every company whose employees use AI runs into this.

Data leaves unnoticed

Customer lists, contracts, personal data and even passwords get pasted into AI tools. The company doesn’t know exactly what has left.

Banning AI doesn’t help

When a company bans AI, people switch to personal accounts and phones. Then nobody is in control.

Nobody wants to be watched

Storing everything people type runs into GDPR and worries employees. A company needs protection, not surveillance.

Example

What the employee sees and what the AI sees

An employee uploaded an article poster and asked the AI what it was about. The answer was correct, even though the AI never saw the author’s name, company or face.

task2mask example, original: a poster for an article on the AI Act and ISO 42001, with the author’s photo and nametask2mask example, anonymised: the same poster after masking. The name, company, website and part of the title are replaced with codes, and the face is pixelatedWhat the employee seesWhat the AI sees
Process

How task2mask works

Every prompt passes through all five steps before it reaches the AI provider.

InterceptThe prompt doesn’t go straight to the AI provider. A browser extension or portal sends it through the company gateway.
MaskNames, account numbers, phone numbers and passwords become codes. The answer comes back with the real data.
ScoreEvery prompt gets a risk score. A secret key or an attempted leak alerts your security team.
Apply rulesThe administrator decides what is allowed and what is blocked, for each AI provider separately.
RetainThe full text and the list of masked data stay with you for investigations and GDPR requests.

Three ways to connect

Choose the connection that fits how your people work with AI, or combine several. All of them share the same rules and log.

Chrome and Safari

Browser extension

Employees stay on their own account and subscription. Personal data is masked before the message leaves the browser.

For company applications

API gateway

Your own tools and SDKs send traffic through a gateway compatible with the OpenAI and Anthropic APIs.

Claude Code, Codex, Cursor

Developer agent

An agent on the developer’s machine masks sensitive data in terminal and code editor traffic.

Control

Security and compliance

task2mask helps you meet GDPR, AI Act and NIS2 requirements and gives you the evidence an auditor will ask for.

Report for the auditor

For a chosen period it produces a Statement of Applicability (SoA) with evidence from the system: what went to the model, what was masked and who revealed what. Export to PDF or JSON.

Employee privacy

We don’t read private conversations. A masked value is revealed only with an admin’s approval, and every reveal is logged.

Repeated incidents

Leak attempts add risk points to the employee’s balance. Past the limit, their access to AI is suspended for a time and they can see why. A course at AI Academy restores access.

GDPR, AI Act and NIS2

Personal data is masked before it reaches the AI, permissions follow role and department, and every action is logged. Xelto is ISO 27001 certified.

Questions and answers

task2mask: FAQ

Does data leave our company?

Not in readable form. The AI provider receives only codes instead of the data. We recognise names and scans locally and process data in a data centre in Poland.

Will the AI still give useful answers if the data is hidden?

Yes. Masking is reversible. The model works with codes, and the employee gets the full answer with the real data back in place.

Do employees have to use different tools?

No. People keep using ChatGPT and Claude. Company data just no longer leaves through them, and you can prove it.

Do you monitor employees’ computers?

No. We monitor only traffic to AI services, not the whole computer. It’s not a traditional data loss prevention (DLP) tool.

Do employees know what is being recorded?

Yes. Every employee has a portal showing who has accessed their data. The Pro plan adds a separate view for the works council.

What if someone asks for their data to be erased (GDPR)?

We delete the data from the records. The erasure itself is logged, and a cryptographic chain still confirms that nobody has altered the other records.

Who pays for using the AI models?

You pay the provider directly for ChatGPT and Claude subscriptions or API usage. task2mask is billed only by the number of users.

What counts as a seat?

Any active AI user, however many devices or channels they use.

What if we go over the limit?

Nothing gets cut off. Once the token limit runs out, only the AI-based layers pause, and pattern-based masking keeps working. If you go over your plan’s seat count, you have 30 days to adjust the plan. Everything keeps working in the meantime, and extra seats are billed only at renewal.

What does deployment look like?

Depending on scope: from a self-service start to deployment run by our team. We set the scope and cost of deployment in the quote.

Can we try task2mask before buying?

Yes, 30 days for up to 10 seats, no card required. We agree the trial details in the quote.

Is there a discount for a longer contract?

Yes. For longer contracts and larger seat counts we offer discounts, and we set the amount in the quote.

Related: Private AI Box, AI agents, Business process automation.