Security

The data is always owned by the platform user. Yandex Cloud uses customer information posted on the platform only to fulfill the purposes of the contract and notifies the client about incidents affecting user data. This is achieved through the:

• principles of secure data processing and data isolation
• use of automated access controls and roles
• use of data encryption
• use of secure data transmission channels
• continuous monitoring and incident responses

By design, LLM does not “remember” specific dialogs, and Yandex Cloud further restricts the storage and use of raw traffic due to privacy, regulatory requirements, threats to the models used, and the cost of infrastructure. In addition, AI Studio has the option of explicitly prohibiting the use of user data.

No, as part of the policy of shared responsibility, Yandex Cloud security specialists implement multiple levels of data isolation.

Stateless inference is a conclusion and a logical inference without a change of state. In machine learning, it means that each model response is performed independently, without taking any previous input or output into account. Key aspects of stateless output are the absence of long-term data storage in memory, independent responses, and simplified implementation.

No, all the available models in Model Gallery and the components of Yandex AI Studio in general are located exclusively in the Russian Federation.

No, we do not store your files. We convert the files into a vector index, and you can manage the lifetime of these vectors and delete them using the corresponding methods in the File Search API.