During the first reporting period, the FAITH project made substantial progress in its technical and scientific work. The project designed and released the initial version of the FAITH AI Trustworthiness Assessment Framework (FAITH AI_TAF), which established a comprehensive methodology for assessing and managing trustworthiness in AI systems. This framework includes a set of over 50 trustworthiness metrics and indicators, forming the basis for risk management and evaluation across the AI lifecycle.
To operationalize this framework, a suite of digital tools and infrastructure was developed and integrated. TrustGuard was created as the system-level orchestrator for trust modeling and risk profiling, allowing stakeholders to track and interpret trust indicators. TrustSense was developed to assess the maturity and readiness of key teams involved in AI development and operations. The AI Model Hub was also implemented, incorporating both the AI Model Passport and Data Passport, which ensure traceability, auditability, and compliance with established data standards. These tools together provide a robust infrastructure for deploying the FAITH AI_TAF across various domains and ensuring adaptability and transparency.
The methodology and supporting tools underwent validation and refinement through workshops and dry-run exercises involving stakeholders from all seven large-scale pilots (LSPs). Feedback from these activities was used to improve requirements and usability, ensuring that the tools met the practical needs of diverse user groups.
The project successfully prepared and initiated seven pilots, each targeting a different critical domain: media, transportation, education, robotics/underwater drones, industrial processes/wastewater management, healthcare, and active ageing. These pilots were set up as real-world testbeds for the FAITH framework and digital tools, with initial activities including the setup of experimental infrastructure, AI system deployment, and integration of trustworthiness assessment mechanisms.
Knowledge and requirements were systematically gathered from each pilot, allowing the FAITH consortium to analyze cross-domain insights and further refine the methodology. This process supported the scalability and adaptability of the overall FAITH ecosystem, enabling it to address the unique challenges and requirements of each sector.
Key achievements during the period include the release of the first operational FAITH AI_TAF with detailed trustworthiness metrics, the development and integration of essential digital tools such as TrustGuard, TrustSense, and the AI Model Hub, and the successful design and launch of seven pilots across critical domains. The project also established a robust technical infrastructure for data and model standardization, promoting transparency, interoperability, and adherence to FAIR principles. All technical deliverables and milestones were met on schedule, with no major deviations or unresolved risks reported during the reporting period.
