Specifications/Technical Information
Thales
Thales
Protect training data without stopping the use of AI.
Encrypt confidential data collected on the Generative AI/RAG platform without modifying the application, and control access to it.
What happens when building an AI platform in-house and training it with confidential data
Internal documents, design specifications, and customer data are gathered in one place on the AI platform.
→More important data is concentrated in one place but is not protected.
The OS and storage administrators can access the contents.
There is no record of who accessed the training data and when.
Three challenges that arise when using confidential data on your own AI platform
Important information such as internal documents, design information, and customer data are consolidated on the AI platform, which means that the impact of any breach would be significant.
Confidential data can be viewed by users with administrator privileges for the OS and storage, increasing the risk of information leaks.
The inability to visualize access to confidential data makes it difficult to detect unauthorized access or insider fraud.
Examples of CTE (CipherTrust Transparent Encryption) usage
Encrypting data that needs to be protected from attacks that steal data through external means, and managing the encryption keys, ensures security when using AI.
Features available at CTE
Because encryption is performed at the file system layer, no modifications to the application or data infrastructure are required.
The OS administrator (root) is separated from access privileges. Policies are applied on a process-by-process basis.
Record references to training data and use them as the basis for AI governance reports.
Inquiry/Document request
In charge of Macnica Thales Co., Ltd.
- TEL：045-476-2010
- E-mail：thales-sales@macnica.co.jp
Weekdays: 9:00-17:00
