Foundations for Risk Assessment of AI in Protecting Fundamental Rights

Risk assessment requires balancing competing rights, not just technical metrics. The framework uses "proportionality analysis" to resolve conflicts between different fundamental rights when AI systems are deployed, recognizing that protecting one right (like security) may sometimes limit another (like privacy).

AI risks must be evaluated in real-world deployment scenarios. Rather than assessing AI systems in isolation, the framework emphasizes analyzing how specific AI applications will actually be used in practice to identify potential legal violations and multi-layered impacts on fundamental rights.

Legal decision-making about AI is inherently flexible and context-dependent. The framework incorporates "defeasible reasoning"—a philosophical approach that acknowledges legal rules can have exceptions and must adapt to specific circumstances—rather than treating AI governance as a rigid checklist.

General Purpose AI systems require broader rights protection analysis. The framework emphasizes that large, versatile AI systems (like large language models) have broader applicability across contexts, making their risk assessment particularly important for protecting fundamental rights at scale.

Bridging theory and practice is essential for effective AI governance. The authors acknowledge their current work is conceptual and commit to developing formal models and practical algorithms, recognizing that responsible AI governance requires translating philosophical frameworks into tools regulators and organizations can actually use.