Knowledge-Augmented Reasoning for EUAIA Compliance and Adversarial Robustness of LLMs

TL;DR

This paper proposes a knowledge-augmented reasoning architecture to help developers and auditors ensure large language models are both compliant with EU regulations and resistant to adversarial attacks, enhancing trustworthiness.

Contribution

It introduces a novel reasoning layer that integrates regulatory compliance and adversarial robustness using knowledge augmentation, bridging legal and technical requirements.

Findings

Demonstrates a functional architecture combining compliance and robustness.

Supports developers and auditors with a reasoning layer based on knowledge augmentation.

Provides a new direction for trustworthy LLM deployment in the EU.

Abstract

The EU AI Act (EUAIA) introduces requirements for AI systems which intersect with the processes required to establish adversarial robustness. However, given the ambiguous language of regulation and the dynamicity of adversarial attacks, developers of systems with highly complex models such as LLMs may find their effort to be duplicated without the assurance of having achieved either compliance or robustness. This paper presents a functional architecture that focuses on bridging the two properties, by introducing components with clear reference to their source. Taking the detection layer recommended by the literature, and the reporting layer required by the law, we aim to support developers and auditors with a reasoning layer based on knowledge augmentation (rules, assurance cases, contextual mappings). Our findings demonstrate a novel direction for ensuring LLMs deployed in the EU are both compliant and adversarially robust, which underpin trustworthiness.