Skip to content

Shadow AI in Ukrainian Companies: Architectural Strategies for Control and Data Leakage Mitigation

Cybersecurity

The rapid proliferation of generative artificial intelligence (AI) tools in employees' daily work creates a new, often invisible threat to Ukrainian enterprises—so-called "Shadow AI." This phenomenon occurs when employees use AI services without official authorization or oversight from the IT department, opening the door to unforeseen risks, including data leakage, intellectual property compromise, and regulatory violations. While the drive for innovation and productivity through accessible AI tools is understandable, it creates tension with the necessity of ensuring data security and legal compliance.

Definition and Prevalence of Shadow AI in Ukraine

"Shadow AI" arises when employees integrate generative AI tools (such as chatbots, content creation tools, and data analysis software) into their work without the knowledge or approval of IT management. These tools can be either public cloud services or local applications. Surveys show that 78% of young Ukrainians have used AI at work without their employer's knowledge, the highest rate among surveyed European countries. Overall, about 80% of employees use AI without established rules, with 34.8% of the information they input into such services containing sensitive data. This practice, while often lacking malicious intent and driven by the desire to work faster and more efficiently, creates significant vulnerabilities.

Key Risks of Shadow AI for Ukrainian Business

Shadow AI carries unique risks that differ from traditional "Shadow IT" due to the specific nature of generative models:

  • Data leakage and confidentiality compromise: Entering sensitive corporate information, personal customer data, or trade secrets into public AI models can lead to their use for training these models, making the data potentially accessible to third parties.
  • Intellectual property infringement: Employees may inadvertently upload proprietary developments or unique business processes, which can lead to their compromise and the loss of competitive advantage. Ukraine already has legislation regarding intellectual property protection, particularly for AI-generated objects.
  • Regulatory and compliance risks: Uncontrolled AI use can lead to violations of the Law of Ukraine "On Personal Data Protection," as well as European regulations such as the GDPR and the EU AI Act. The latter has extraterritorial effect, and Ukrainian companies offering AI systems in the EU market or whose results are used in the EU are subject to its requirements. Fines for violating the EU AI Act can reach up to 35 million euros or 7% of global annual turnover.
  • Inaccurate or biased results ("hallucinations"): AI models can generate false or biased information, which can lead to erroneous business decisions, reputational losses, and legal issues, especially in areas requiring high precision, such as legal advice.
  • Security vulnerabilities: Risks include "prompt injection" attacks, where attackers manipulate AI systems by entering malicious queries.

Architectural Strategies for Control and Risk Mitigation

To effectively manage Shadow AI risks, a comprehensive architectural approach that combines technical controls and organizational policies is necessary.

It is important to set clear boundaries for the use of AI tools, especially those that process confidential data. This may include deploying internal, controlled AI solutions for processing sensitive information, rather than using public external services. Internal AI platforms provide greater control over data, its storage, and its use for model training, whereas external services, while convenient, carry risks of data transfer to third parties.

Applying the principles of "Privacy by Design & by Default" at all stages of AI system integration is critical. This means that data protection must be embedded into the system architecture from the very beginning, rather than added as an afterthought.

Checklist for identifying and assessing Shadow AI risks:

  • Do employees use unauthorized AI tools to perform work tasks?
  • Is confidential information (personal data, trade secrets, IP) being uploaded to public AI services?
  • Are there network traffic monitoring mechanisms to detect unauthorized access to AI platforms?
  • Is there a regular audit of AI tool usage within the company?
  • Do employees understand the potential risks associated with data leakage via AI?
  • Are there clear rules regarding the use of AI to create or modify published content?
  • Are the risks of AI "hallucinations" assessed for critical business processes?

Developing a Corporate AI Usage Policy

The absence of a clear AI usage policy turns it into a zone of legal, financial, and reputational threats. Developing and implementing a comprehensive corporate policy is a fundamental step.

Such a policy should include:

  1. Data classification: Defining which types of data (public, internal, confidential, personal) can be entered into different categories of AI tools (internal, approved external, prohibited).
  2. List of approved tools: Creating a list of permitted AI services that have undergone security and compliance assessment.
  3. Responsible usage rules: Instructions on how employees should interact with AI, including fact-checking, disclosing AI use in content creation, and prohibiting the entry of sensitive data into unauthorized systems.
  4. Training and awareness: Regular training for employees on Shadow AI risks, company policy, and best practices for safe AI use.
  5. Monitoring and audit mechanisms: Implementing systems to track AI tool usage, detect violations, and respond to incidents.
  6. Accountability: Clearly defining responsibility for policy violations, including updates to employment contracts and contractor agreements.

The Path to Secure and Managed AI Integration

Ukrainian enterprises are on the verge of significant changes, where AI can become a powerful growth catalyst. However, without proactive management of Shadow AI risks, these opportunities could turn into serious threats. IT department heads, CISOs, and CIOs must not only realize the scale of the problem but also actively implement architectural strategies and policies that ensure safe and responsible AI integration. This will allow not only for the protection of critical company assets but also for the creation of a culture of innovation where AI is used as a tool for development, not a source of uncontrolled risk. Harmonization with European AI regulation is inevitable, and companies that start preparing now will secure a competitive advantage and resilience in the future digital landscape.

Sources

  1. 01cms.lawAI laws and regulations in Ukraine | CMS Expert Guide
  2. 02lexology.comWhat Ukraine's AI Law Could Look Like
  3. 03pwc.comUkraine AI Regulation Overview
  4. 04iso.orgISO/IEC 42001:2023

Does any of this match your situation?

Tell us where you are now — we will suggest a practical route for infrastructure, cloud or security.

Discuss a project