Security and data protection
We design AI and data security together.
We decide together from the start which data is processed where, and build the architecture around KVKK, GDPR and your internal policies.
Deployment options
Three models, matched to the sensitivity of your data
For each workflow we recommend the best option based on the type of data. Most organisations use all three together.
Tools under enterprise agreements
We connect the enterprise editions of Copilot, ChatGPT, Claude and Gemini to your identity system and verify data retention and training settings during deployment.
Best for: General office work, correspondence, summarisation, research
Sensitive data inside, general work in the cloud
We build a layer that routes personal or commercially sensitive data to in-house models and general work to cloud tools. Users see a single interface.
Best for: Organisations handling several classes of data
A deployment that keeps data inside
We run open models such as DeepSeek, Llama and Mistral on your own servers with vLLM and Open WebUI. Data is processed within your network.
Best for: Finance, healthcare, public sector, defence, trade secrets
In every project
Security is part of the architecture from day one
We start with data classification
In the first week of the project we decide together which data can be processed with which tool, and where.
Identity and access management
Single sign-on (SSO), role-based access and users reaching only the documents they are authorised to see.
Logging and traceability
Logs and reports that answer who used which tool with which data.
Retention and deletion rules
Retention periods for chat history, uploaded documents and vector databases that follow your policy.
KVKK and GDPR assessment
We address transparency, consent and cross-border transfer with your legal team and design the architecture accordingly.
Usage policy and training
An AI usage policy for employees and hands-on training on which data can be shared where.
Frequently asked
Security and data protection questions
KVKK is Türkiye's personal data protection law. For projects in the EU we take GDPR requirements into account in the same way.
Will our data be used to train AI models?
In enterprise editions, providers contractually commit to keeping customer data out of model training. During deployment we verify these settings and retention periods one by one. In an on-premise deployment, data is processed within your own network.
Can we use cloud-based AI tools under KVKK and GDPR?
It depends on the type of data and the conditions for cross-border transfer. We first map your data inventory and clarify which work can run in the cloud and which should stay inside. The final legal assessment is made together with your legal team.
What hardware does an on-premise deployment need?
It depends on the number of users, the model and the expected speed. During evaluation we measure the need and show in numbers whether to proceed with your existing servers or a new GPU investment.
How do you work with our information security team?
We submit the architecture, access model and logging design for your information security team's approval and integrate with your existing identity, logging and monitoring systems. At the end of the project all configuration and documentation is handed over to you.
Let's plan together where your data is processed
In a free discovery call we listen to your data landscape and discuss which work belongs in the cloud and which should stay inside.