AccountabilityTransparencyTrustworthinesscs.AI
TAI Scan Tool: A RAG-Based Tool With Minimalistic Input for Trustworthy AI Self-Assessment
Athanasios Davvetas, Xenia Ziouvelou, Ypatia Dami, Alexios Kaponis, Konstantina Giouvanopoulou, Michael Papademas
July 23, 2025
TRUST-AI@ECAI
1 citations
1 influential
Key Takeaways
-
The tool automates AI compliance assessment: TAI Scan Tool uses artificial intelligence to help organizations determine whether their AI systems comply with the EU AI Act, reducing the need for manual legal review and making compliance more accessible to non-legal experts.
-
It simplifies the compliance process through "minimalistic input": Rather than requiring extensive documentation, the tool uses a two-step approach (pre-screening and assessment) designed to work with limited information, making it practical for organizations with varying levels of compliance expertise.
-
The tool correctly identifies risk levels: According to the paper's evaluation, the system successfully predicted risk levels across different AI system types and retrieved relevant legal articles, suggesting it can reliably guide organizations on their regulatory obligations under the AI Act.
-
High-risk systems receive particular attention: The tool's reasoning focuses heavily on comparing systems to "high-risk" categories defined in the AI Act, which the authors note is appropriate since these systems require the most careful consideration and regulatory scrutiny.
-
It bridges the gap between AI developers and legal requirements: By automatically retrieving relevant AI Act articles alongside risk assessments, the tool helps non-legal audiences understand not just what their compliance obligations are, but which specific regulations apply to their systems.