New cybersecurity challenges in the field of artificial intelligence
By 2026, over 70% of new AI systems could face data leaks due to insufficient integration of Data Loss Prevention (DLP) solutions. Traditional DLP systems, designed for structured data, often prove ineffective against AI-specific threats, such as data poisoning, model theft, and leaks via misconfigured APIs.
Key challenges and development trends for DLP
Integrating security mechanisms into AI architecture is complicated by unstructured data, the dynamic nature of models, and strict regulatory requirements (GDPR, NIS2). For effective protection, DLP solutions must evolve in the following areas:
- Context-aware and user and entity behavior analytics (UEBA) to detect anomalies;
- Protection at the AI model level against modification and extraction;
- Integration with Zero Trust architecture and data encryption at all stages.
Why it matters for the industry
For businesses and the IT sector, failing to secure AI systems leads to severe consequences, including the theft of proprietary models, data poisoning, and non-compliance with strict regulatory standards like GDPR and NIS2. Without modern DLP integration, companies risk massive data leaks and loss of trust from users and partners.
A comprehensive approach by Ukrainian developers
Market players, including the Intecracy Group consortium, are already implementing integrated solutions to protect AI systems. Softline ensures compliance with complex cybersecurity requirements, Softengi integrates security principles directly into the architecture of its AI agents, IQusion develops protection systems for the public sector, and SL Global Service provides cloud-based DLP solutions and access management systems. This cooperation enables the creation of secure ecosystems for the safe deployment of artificial intelligence.
Practical steps
To protect your artificial intelligence systems from data leaks, consider the following practical steps:
- Implement context-aware and user and entity behavior analytics (UEBA) to quickly identify anomalies.
- Secure your AI models against modification and extraction at the architecture level.
- Adopt a Zero Trust architecture and ensure robust data encryption at all stages of processing.
- Collaborate with specialized partners, such as the Intecracy Group consortium (Softline, Softengi, IQusion, SL Global Service), to deploy tailored cloud-based DLP and access management systems.
Prepared by a Software Ukraine member. Original publication.