Security and KVKK

Data does not leave the company. What stands behind that sentence?

Where each piece of data sits, what the agent can access, what is logged, and who holds which role under KVKK (data protection law): one by one.

A team reviewing records and documents
A team reviewing records and documents
A team reviewing records and documents
In short

Panda AI and Panda Tuval are installed on the customer's own server or in the customer's own cloud account; no record such as accounts, prices, payroll or quotes is sent outside the company boundary. The agent accesses the ERP only as far as the permissions written during setup, every action is written to the audit log and every out-of-rule record goes to human approval.

The rules behind that sentence

Where does the data sit?

Where does the data sit?

Setup is done where you choose: your company server, a cloud account opened in your name or the environment where your ERP runs. The documents the agent reads, the records it matches and the reports it produces stay in this environment. Panda has no central data pool; support access is through a channel you open, for a period you set.

  • Company server or your own cloud account
  • The ERP database is not copied, it is read in place
  • Support access through a channel you open and for a period you set
What does the agent access?

What does the agent access?

The access scope is written together during setup: which table gets read access and which gets write access. The pilot starts at read-only or draft level (L2); the agent can open a record directly only once you write the rule and widen the permission. Permissions can be narrowed at any time; a request outside the scope technically cannot be made. It does not go to approval, it never occurs.

  • Read/write permission per table
  • Autonomy levels L1–L4, which do not widen without your approval
  • Permission changes are in the audit log too
What is recorded?

What is recorded?

Every action is written to the audit log: which document was read, which record it was matched with, what was done under which rule, who approved and when. It can be queried and reported retroactively. You can see why a record was opened or why it went to approval in two clicks; in an audit, the answer to "who did this" is ready.

  • Document → record → rule → approval chain
  • Reversible actions
  • Report can be exported
Does the AI model see the data?

Does the AI model see the data?

In the step of reading and matching documents, the model runs in the environment set during setup; which model runs where, and which fields are never sent outside the company (for example payroll or a price list), is written with you during setup. Your data is not used to train models. The details are discussed with your technical team in the discovery call; what applies here is not a promise but the rule written in the setup.

  • Sensitive fields are excluded during setup
  • Models are not trained on your data
  • Environment and model choice is a setup decision
Who has access, and how?

Who has access, and how?

Users work with their roles in the ERP; the agent does not require you to learn a separate screen. The approval screen comes inside the ERP or by WhatsApp/email; approval authority is defined per person. Access for the Panda support team is limited to the account and the period you open; it closes when the period ends.

  • Role-based approval authority
  • Support access is time-limited and logged
  • When an employee leaves, access closes together with the ERP
Backup and continuity

Backup and continuity

Because the agent is installed next to your ERP, your ERP's backup routine does not change. The agent's own configuration (flows, rules, audit log) is backed up in your environment. If the agent stops, work does not stop: the process carries on by hand as it did before the agent, and the documents that fall into the queue are processed when the agent is back.

  • The ERP backup routine does not change
  • Flows and the audit log are backed up in your environment
  • If the agent stops, the process continues by hand
Roles under KVKK

Roles under KVKK

For records containing personal data, you are the data controller; Panda acts on your instructions within setup and support. The data processing agreement, retention periods and privacy notices are adapted to your KVKK (Turkey's data protection law) setup. Which fields count as personal data and how they are processed is determined together with your legal adviser; we build the technical side.

  • Data controller: you
  • Data processing agreement during setup
  • Retention period and deletion rule are defined

See how Security and KVKK would work in your business

Tell us briefly about your process and we will walk through the screens with your own example.

  • Reply the same business day
  • 15+ years of experience
  • 90 client businesses

Frequently asked questions about Security and KVKK

Can Panda see our data?

Only through the support access you open, for the period you set. There is no central Panda data pool; the setup is in your environment.

What happens to the agent if the internet drops?

On a company server, it keeps working on the same network as the ERP. In a setup in your cloud account, documents fall into a queue and are processed when the connection returns; the process continues by hand.

Can we narrow the agent's permissions later?

Yes, at any time. Permissions are defined per table and any change is written to the audit log. You do not need Panda to narrow them.

Do you have ISO certification or an independent audit report?

On this page we describe only the rules that are written during setup and can be verified. Send your request for a certificate or audit document in the discovery call; we will share the current status openly.

Can we start with a process that has no personal data?

Yes, and most pilots start that way: in processes such as the morning report, quote comparison and stock gap, personal data is either absent or excluded.

Let us talk about the process first, then the software

Describe your process in a few sentences and we will show you the modules that fit.

  • 15+ years of enterprise software experience
  • In the cloud or on your own server
  • Processes compliant with KVKK (Turkey's data protection law)