Scholarlink

Large-scale generative AI models such as ChatGPT perform well in text generation, automatic dialogue, and multimodal tasks. However, the widespread deployment of these systems presents challenges related to their credibility and trustworthiness, encompassing issues such as security vulnerabilities, interpretability limitations, reliability concerns, and privacy risks. This study systematically examines the latest advancements in dealing with these challenges through a comprehensive, interdisciplinary approach. By integrating insights from technology, ethics, and policy domains, the research underscores the multifaceted implications of deploying large-scale generative AI systems. In addition, this paper provides practical insight into a structured framework centered around the core pillars of trusted AI systems, guiding the design, development, and deployment of generative AI models that meet societal expectations, helping to improve understanding of the credibility of generative AI, and promoting its responsible integration in various applications.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Survey on Trustworthy Systems in ChatGPT-Like Large-Scale Generative AI Models: Security, Interpretability, Reliability, and Privacy Considerations

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Dun Li,
  • Hongzhi Li,
  • Jiatao Li,
  • Noel Crespi,
  • Roberto Minerva,
  • Kuan-Ching Li,
  • Lingxiang Hu,
  • Wenhao Shao

摘要

Large-scale generative AI models such as ChatGPT perform well in text generation, automatic dialogue, and multimodal tasks. However, the widespread deployment of these systems presents challenges related to their credibility and trustworthiness, encompassing issues such as security vulnerabilities, interpretability limitations, reliability concerns, and privacy risks. This study systematically examines the latest advancements in dealing with these challenges through a comprehensive, interdisciplinary approach. By integrating insights from technology, ethics, and policy domains, the research underscores the multifaceted implications of deploying large-scale generative AI systems. In addition, this paper provides practical insight into a structured framework centered around the core pillars of trusted AI systems, guiding the design, development, and deployment of generative AI models that meet societal expectations, helping to improve understanding of the credibility of generative AI, and promoting its responsible integration in various applications.