Capabilities and Challenges of AI in Digital Governance by Stamatis Ezovalis

AI has roots back to the 1950s but has seen explosive growth recently, particularly with the 2022 release of ChatGPT and NVIDIA’s dominance in 2024

Current AI models are moving beyond simple task execution to learning and generating content through neural networks, deep learning, and cognitive algorithms

Recent testing of prominent AI models (Meta, ChatGPT, Google) against AI Act requirements showed concerning compliance rates:

• Meta scored 42%
• JCPT scored 49%
• Google scored 85%

Key challenges in AI governance include:

• Cybersecurity vulnerabilities
• Bias in algorithms
• Prompt hijacking risks
• Privacy concerns
• Need for reliability and transparency

Practical public sector applications include:

• HR allocation in schools
• Ministry of Education AI assistance
• Fraud detection in procurement
• Healthcare prescription management
• Document processing automation

Trust framework implementation is crucial for AI governance, focusing on:

• Privacy
• Reliability
• Fairness
• Sustainability
• Transparency

Despite AI’s capabilities, there are still limitations in accuracy and detail recognition, as demonstrated in image generation tests

Integration of AI in government services requires careful consideration of responsibility and ethical implications while balancing innovation with public trust