Security is not a feature we add later — it is the reason the platform exists. Below is how the platform is built and what we hand to your security team.
| ALDIS | Public cloud LLMs | Global enterprise suites | |
|---|---|---|---|
| Data never leaves the client's premises | ✓ | ✕ | – |
| Native Kazakh & Russian (plus Arabic) | ✓ | – | ✕ |
| No dependence on foreign cloud / LLM APIs | ✓ | ✕ | ✕ |
| Integrates with 1C and local reporting standards | ✓ | ✕ | – |
| Custom-tuned per client, deployed in weeks | ✓ | ✕ | – |
| Cost | Lower | Per-token | High |
The Law on Artificial Intelligence (No. 230-VIII) adds AI-specific duties: risk classification documented in a system passport, informing users that they are dealing with AI, human involvement in decisions with legal effect, no prohibited capabilities, and periodic review of the risk passport.
No. Models run in your data centre or on dedicated servers, inference does not depend on internet access, and nothing is sent to third-party AI providers. Data residency is enforced by the architecture, not by a contract clause.
The usual blocker is an external LLM API, which does not fit the requirements on placement, access, logging and cryptographic protection. An on-premise deployment is designed to meet them, and we hand over a conformity matrix so your security team can verify each control before acceptance.
Storage and transport are encrypted (AES-256, TLS 1.2+); integration with certified cryptographic protection is available on request.
Personal data are masked before the model processes them, retention and deletion are configurable, and customer data are separated per installation with no cross-client training.
A 30-minute demo on your use case: finance, legal, HR or customer service.