Cloud provider Cloud.ru has expanded the capabilities of its Foundation Models service, providing access to more than 50 foreign large language models through a single API. The update includes the integration of security tools to protect corporate data when using global neural networks.

What Happened
The Cloud.ru Foundation Models catalog now includes model families from Alibaba, DeepSeek (including V4-Pro and R1), Z.ai (GLM-5.2), MiniMax, and GigaChat. Access to these models is provided through a Unified API. To ensure query security, the Guardrails tool has been integrated, which automatically masks sensitive information within the corporate perimeter before it is sent to external systems.
Context
Previously, Russian companies faced infrastructural and legal challenges when attempting to use top global models directly, including difficulties with payments, integrating different APIs, and complying with regulatory requirements. This update effectively creates a local technological hub, similar to OpenRouter, but adapted to the Russian market specifics.
Why It Matters for the Industry
For the industry, this represents the creation of an important abstraction layer and a secure AI Gateway. The service allows companies to use cutting-edge SOTA models while complying with Federal Law No. 152-FZ, and reduces the Time-to-Market for new AI products. This simplifies the scaling of multi-model pipelines in the enterprise sector and reduces operational risks when working with cross-border APIs.
Why It Matters for Users
Developers and business users gain the ability to switch between specialized neural networks within a single cloud provider: for example, using GLM-5.2 for coding tasks, DeepSeek R1 for complex reasoning, or the Gemini family for long-context tasks. Simultaneously, the need to independently resolve complex integration issues and problems with making payments for foreign APIs is eliminated.
What Is Not Yet Known / Limitations
Despite the use of Guardrails for data masking, cross-border legal risks remain when using external APIs, which are not completely mitigated even with the presence of data protection tools.
Sources
Author
Look at AI, Editorial Team
