The integration of artificial intelligence (AI) in business operations presents both opportunities and challenges, particularly concerning the management of sensitive information. Companies face the risk of losing their competitive advantage if unpublished research results, pricing strategies, client data, or contract details are inadvertently disclosed. This necessitates a precise oversight of who can access confidential company data during AI processing.
Currently, partnerships with cloud service providers rely on a blend of technical security measures and trust-based agreements, including non-disclosure agreements (NDAs) and formal security protocols. As cloud resource processing becomes more dynamic, mechanisms that technically limit and verify data access are gaining significance. Solutions that physically restrict access to protected information from privileged infrastructure components are increasingly essential, moving beyond reliance on legal declarations alone.
Data protection during processing is typically ensured through encryption of data stored on disks and transmitted over networks. However, this approach does not eliminate all risks associated with active data usage. In standard AI applications, processing encrypted information requires access to its unencrypted form, heightening the importance of securing the environment where computations occur. AI-oriented architectures must safeguard resources throughout their lifecycle, from the moment of decrypting input information, through the analytical process, and generating responses, to re-encrypting data before logging or backup.
A significant technological advancement in data protection during use is Confidential Computing, which allows for the isolation of computational processes and the data being processed from the operating system, hypervisor, other applications, and even the cloud infrastructure operator. This represents a crucial shift in the security approach, which traditionally relied heavily on trust in the provider’s procedures, agreements, and organizational safeguards. Confidential Computing enhances these mechanisms with hardware-level isolation.
Another critical aspect of information security is the management of cryptographic keys that grant access to encrypted resources. For cloud-based companies, this involves determining who can use these keys and how to restrict access from individuals and systems managing the infrastructure. Hardware Security Modules (HSM) are often used to secure key management, although maintaining such modules incurs costs related to procurement, integration, and infrastructure management.
In some cases, virtual solutions offer flexible cloud-based key management and cryptographic operations without the need for physical hardware investment. One such solution is Oktawave Trust Vault, developed under the Next Gen Cloud initiative. This platform utilizes Intel SGX technology to protect keys and cryptographic operations while combining centralized management functionality with scalable cloud resources. This model can be a cost-effective alternative to maintaining dedicated HSM devices.
Furthermore, the sovereignty of data encompasses more than just server location for businesses utilizing AI; it also entails maintaining control over strategic information, minimizing dependency on external vendors, and mitigating risks of unauthorized access to confidential resources. This perspective extends beyond merely where servers are located to include who manages the infrastructure, their permissions, the principles of data processing, and whether a business can securely migrate data to a different environment.
The European initiative IPCEI-CIS (Important Project of Common European Interest – Cloud Infrastructure and Services) supports the development of solutions in this area. Among the 19 participating companies is Oktawave, which is executing the Next Gen Cloud project aimed at enhancing secure data processing technologies and cloud environment interoperability, particularly for advanced analytics and AI.
The Oktawave Trust Vault is part of a broader research and development project, Next Gen Cloud, conducted within the framework of the European program IPCEI-CIS. Additionally, the Data Fabric platform is being developed to unify access to distributed storage resources and automate information lifecycle management processes. This includes authorization and authentication mechanisms in multi-cloud environments, allowing seamless collaboration of services across different clouds and edge locations.
In addressing challenges related to cloud infrastructure, Trust Vault focuses on protecting keys and cryptographic operations, Data Fabric streamlines management processes for distributed resources, and interoperability mechanisms reduce barriers between environments. The common objective is to develop an infrastructure where security, data access, and collaborative capabilities of various platforms are integrated from the architectural level.
Concerns over uncontrolled disclosure of business secrets need not delay the deployment of AI algorithms. Organizations can achieve a balance between analytical efficiency and information resource protection, provided they complement legal safeguards with technical control mechanisms embedded directly in system architecture. The key to utilizing AI securely lies in implementing solutions that protect data throughout its entire lifecycle, particularly during active processing phases. These decisions will ultimately determine the ability to leverage the potential of artificial intelligence without unnecessarily risking the company’s most valuable resources.
