人工智能开源模型训练的伦理风险

中图分类号：B82-057 文献标识码：A 文章编号：2095-6916（2025）13-0009-04

Ethical Risks in the Training of Open-Source Artificial Intelligence（AI） Models

Yan MingLong Lida （JilinNormalUniversity，Siping136000）

Abstract：From a techno-ethical perspective，large AI models exhibit two distinct development models，open Π-Π source and closedsource，eachemonstratinguqueadantagesandnherentiitatios.Troughsystematicnalysisofeticalisksintetraiingofe sourceAImodels，tisstudyidentifiessignificantchallngesinareassuchasdataprivacymodelatacks，technologicalmonopliesand algorithmicdiscriminationAlthoughoensourcemodelspromotetechnologicalinovation，issuessuchasunauthorizeddatausealiciouscodeinjectioandalgoriticdisriiationayexacerbateocialijusticeThisdypropossacollborativegoveaefa work，emphasiingteanementofrasparencyataqualityssuraneandericipleofticalpririadvocatingforulti stakeholder co-governance to balance technological openness and risk prevention.

Keywords：AI;open-source models；ethical risks；data privacy；algorithmic discrimination

随着人工智能技术的飞速发展，人工智能开源大模型的训练逐渐成为行业内部技术升级的主流趋势，开源模式的开放性和透明性使得人工智能技术获得了更广泛的应用和技术创新，但开源模型训练在推动技术进步的同时引发了大量的伦理风险和挑战。（剩余5557字）