Viaz is a platform that organizes the data in your company’s documents and decides what of it may reach AI tools. Our own ML model classifies documents: it recognizes the document type as well as personal, financial and medical data, including in scans and archives. On that basis, Viaz shows excessive access, documents past their retention period and duplicates, while the Secure AI module controls which data employees send to Copilot, ChatGPT and other tools. It runs entirely in your infrastructure.

Secure AI checks the prompts and files employees send to AI tools before they leave the company. You set policies for user groups and tools: allow, mask sensitive data (for example, a PESEL number is replaced with a token) or block sending. Viaz recognizes files by their content among the documents it has already scanned, so it can block a document under legal hold or marked as high risk, even if it has been renamed. Every use is logged.

All of them, regardless of how an employee uses them: chats in the browser (ChatGPT, Claude, Gemini, Copilot and others), API calls from applications and scripts, command-line tools (CLI) and AI built into applications. Depending on the channel, a browser extension, a gateway (proxy) for APIs or a workstation agent does the work.

Yes, for two reasons. First, in Viaz you set the list of approved AI tools: since Copilot is your company tool, you can block sending data to other chats that employees use on their own, or allow it only after the data is masked. Second, Copilot sees everything the user has access to. It does not give anyone new access, but it uses existing access in seconds, including access that is too broad. Viaz shows what Copilot can see in SharePoint and OneDrive, and which access is worth revoking.

Viaz does not replace Purview, it complements it. On an E3 license, Purview means manual labeling: an employee applies labels file by file, so it mainly covers new documents, while archives and scans stay unlabeled. Viaz automatically classifies your entire historical estate, including scans, and can write the result as Microsoft (MIP) labels in your taxonomy. A quick check: what label coverage do you have on documents older than a year? On E5, Purview classifies automatically, but mainly in the Microsoft cloud. With a single tool, Viaz also covers file servers and scans, and on top of that detects duplicates, which Purview does not show.

No. DLP acts at the moment data leaves the company. Viaz organizes data earlier: it classifies documents and shows where they are, so DLP gets labels and has something to work with. Most companies we talk to have DLP, and that is exactly why they need classification underneath.

Yes. It recognizes Polish identifiers (PESEL with validation, NIP, REGON, ID card number, IBAN) and identifies the document type by its content, not its file name. Scans and images go through OCR.

No. The model ships ready-made and runs on your side. Your documents are never used for training and never reach the vendor.

We deploy Viaz on-premise, in a private cloud or hybrid. Data, classification results and logs stay with you; classification runs on our own model, without sending documents to external services. Repository scanning (SharePoint, OneDrive, file servers, email) runs without agents. Secure AI works through a browser extension, so it does not need an agent either. Protection can be extended with a workstation agent that analyzes traffic on the user’s computer, including from desktop applications. Viaz makes no connections to the vendor’s infrastructure.

A few hours. We install Viaz in your infrastructure, connect the first data source and configure sign-in through Entra ID. The first scan starts the same day, and the first results are visible within 24 hours. On your side, you need someone from IT to prepare the environment and grant permissions to the sources.

Read-only by default. A scan only needs read access to files and to permissions in the sources, so Viaz changes nothing. You grant write permissions only when you want Viaz to write labels on files or carry out approved actions, such as narrowing access or deleting a document after its retention period. We agree the scope individually during deployment.

No. A scan does not affect users’ work in SharePoint or on file servers. The first scan goes through the entire estate, and subsequent ones are incremental: they cover only files changed since the previous scan.

Microsoft 365: SharePoint Online, OneDrive, Teams, Exchange Online (mailboxes and attachments) and Exchange archives. Exchange Server in your own infrastructure (on-premise). File servers and network drives (SMB shares), plus FTP servers. Google Workspace: Google Drive. Azure: Azure Files, Azure Blob Storage and Azure Tables. Amazon Web Services: Amazon S3. Databases: SQL Server, PostgreSQL, Oracle and MySQL / MariaDB. Scans and images in each of these sources go through OCR.

Yes, sign-in works through Microsoft Entra ID (SSO), so you grant and revoke access to Viaz the same way as for other company applications. Permissions in the console are assigned to roles, and actions such as deleting files or narrowing access are approved by designated people, for example from compliance or IT.

Yes, the Secure AI log contains the prompt content, the detected data, the rule that fired, the user, the tool and the time. This lets you show which data reached AI. The goal is to protect company data, not to assess people’s work, but employees must be informed about it. Only people with the appropriate role in the console can access the log, and the rules for using it are worth describing in your AI policy together with your DPO.

Viaz prioritizes by risk and proposes a specific action: revoke the link, restrict to a group, delete after retention. The data owner approves, Viaz executes and logs it. We start with a narrow scope so the first results can be handled within weeks. M3M supports the review and decisions.

Yes. Deletion is a soft delete that can be restored, and access restrictions can be rolled back from the audit log. Every action has an approver and a date.

Support is provided directly by the maker of Viaz, under an SLA with agreed response times. Support is in Polish, and together with updates it is included in the license price.

Two paths. Free permissions review: 2–4 weeks, one source, read-only, with a report at the end. Paid pilot of the full platform: installation on your side, a scan of the agreed scope, Secure AI for a group of users, and a closing report; the cost is credited toward the license. The annual license includes deployment, updates and support in Polish.

Viaz runs in production at an international restaurant chain operator with several thousand employees. In the financial sector we are running proposals and pilots in banking, insurance and brokerage. The team behind the product serves as data protection officer for, among others, a brokerage house. We can arrange a reference call on request.