支持AI系统伦理与监管认证的框架
人工智能
2025-10-02 v1 计算机与社会
数据库
摘要
人工智能正在迅速成为支柱技术,显著影响欧洲的社会和经济景观。然而,AI的扩散也引发了一系列关键的伦理、法律和监管挑战。CERTAIN(Certification for Ethical and Regulatory Transparency in Artificial Intelligence)项目通过开发一个综合性框架来解决这些问题,该框架整合了监管合规、伦理标准和透明度。本文是一份position paper,概述了构建框架核心组件的方法论步骤。具体而言,我们提出:(i)用于结构化AI生命周期管理的语义机器学习运营(MLOps),(ii)用于确保可追溯性和可责任性的本体驱动数据血缘跟踪,(iii)用于实现监管要求的监管运营(RegOps)工作流程。通过在多样化的试点项目中实现和验证其解决方案,CERTAIN旨在推动监管合规,并促进符合欧洲标准的负责任AI创新。
引用
@article{arxiv.2510.00084,
title = {Towards a Framework for Supporting the Ethical and Regulatory Certification of AI Systems},
author = {Fabian Kovac and Sebastian Neumaier and Timea Pahi and Torsten Priebe and Rafael Rodrigues and Dimitrios Christodoulou and Maxime Cordy and Sylvain Kubler and Ali Kordia and Georgios Pitsiladis and John Soldatos and Petros Zervoudakis},
journal= {arXiv preprint arXiv:2510.00084},
year = {2025}
}
备注
Accepted for publication in the proceedings of the Workshop on AI Certification, Fairness and Regulations, co-located with the Austrian Symposium on AI and Vision (AIRoV 2025)