AI4I’s AI Security (AIS) Lab has secured its first industrial partnership with HikmaAI, marking a major milestone in the Institute’s strategy to bridge the gap between advanced research and real-world industrial applications.
The collaboration tackles a critical roadblock in enterprise AI adoption: rigorously evaluating model robustness against adversarial attacks.
To achieve this, the joint research teams developed a specialized multimodal dataset featuring thousands of image-text and audio attacks, designed to stress-test the resilience of AI agents in production-grade environments.
This dataset equips HikmaAI with a strategic asset to audit safety, enhance reliability, and accelerate the development of highly secure AI solutions.
Completed within a three-month timeframe, the project leveraged High-Performance Computing (HPC) infrastructure from the IT4LIA AI Factory via the EuroHPC initiative, enabling the large-scale parallelization of attack generation.
This inaugural partnership underscores the role of AI4I’s R&D&E Labs as catalysts for applied research and technology transfer, fast-tracking the adoption of trustworthy artificial intelligence across the industrial sector.
