As Artificial Intelligence (AI) systems, mostly deep neural networks, become increasingly important in critical and everyday applications, it is crucial to make them more explainable, secure, and trustworthy. Traditional approaches to AI development often prioritize performance over transparency or robustness, resulting in models that remain difficult to interpret and vulnerable to misuse. That is why the special session Explainability and Security in Trustworthy Artificial Intelligence Systems aims to address these challenges by bringing together researchers and practitioners at the intersection of explainable AI (XAI), AI security, and AI for security.

The session aims to explore how XAI can enhance the usability of AI systems, enabling human stakeholders to make more informed decisions and thereby improving society's trust in such systems. The session will also focus on advances in methods for securing AI against security threats, such as adversarial threats, manipulation, and systemic vulnerabilities. Recognizing that robustness is essential for reliable deployment, it will also emphasize that XAI and security should be perceived as inseparable. The topics of interest include using AI algorithms (for example, Large Language Models) to support explainability and security of other AI algorithms in a cross-model way. 

Moreover, the session aims to look beyond solutions created solely for AI algorithms by recognizing that intelligent systems have a dual role, both as a subject needing explainability and security, and as a tool to strengthen the transparency and security of other systems that integrate AI alongside other critical components. In such applications, AI can be used for tasks such as anomaly detection, monitoring, and reasoning. By fostering dialogue across these complementary perspectives, the goal of the special session is to support the design of intelligent systems that serve human needs, while also promoting resilience and transparency in increasingly complex intelligent ecosystems.