In the heart of the Greek state, behind the screens of public administration offices, a silent but dangerous revolution is unfolding. The term "Shadow AI" describes the use of generative AI tools, such as ChatGPT or Claude, by employees to perform their duties without the approval, supervision, or knowledge of IT departments. While the intention is often to improve productivity, the risks to national security and personal data protection are now spiraling out of control.
The Efficiency Paradox
Greek public administration, traditionally associated with bureaucracy and slow processes, has found an unexpected ally in Artificial Intelligence. Employees tasked with drafting lengthy reports, translating documents, or analyzing data are turning to free AI tools to save time. However, this "private initiative" occurs outside of security protocols. When an employee inputs a confidential government document into a public AI model for summarization, that data is no longer private. It becomes part of the model's training set, making it potentially accessible to third parties or exposed to cyberattacks.
Data Risks and the Shadow of GDPR
With the EU AI Act now in full effect in 2026, Greece faces serious legal challenges. The use of unauthorized AI systems in the public sector violates fundamental principles of the GDPR and the new European framework. The problem is not just data leaks, but also the "illusion of validity." AI models often produce "hallucinations," presenting false information as fact. If a public official bases a decision or a bill on inaccurate data generated by AI, the consequences for citizens and the rule of law could be catastrophic.
The Lack of Central Strategy
Despite digital transformation efforts in recent years, there appears to be a massive gap in employee training and guidance. The state has yet to provide a secure, "closed" AI environment for internal use. The absence of official tools forces the more tech-savvy employees to become "shadow users." The solution is not prohibition—which is practically impossible to enforce—but the establishment of strict usage frameworks. We urgently need a "Government AI Cloud" where operations are performed with guarantees of privacy and ethical standards.
Toward Controlled Integration
To combat the Shadow AI phenomenon, the Greek government must move along three axes: education, infrastructure, and regulatory compliance. Employees must understand that AI is not a simple search engine but a data processor with a memory. Simultaneously, the creation of specialized AI Task Forces in every ministry is essential for evaluating tools. Artificial Intelligence can skyrocket public sector efficiency, provided it is not left to chance and the individual judgment of each employee.
