[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"tag-ai-安全":3},{"tag":4,"articles":11,"peer_article_count":12},{"id":5,"name":6,"slug":7,"article_count":8,"description_zh":9,"description_en":10},"886ee003-785a-4a10-8b24-3d15e1c090cc","AI 安全","ai-安全",5,"AI 安全涵蓋模型濫用、供應鏈風險、權限控管與治理設計，從漏洞挖掘、惡意倉庫到審批閘門都在討論範圍內。對開發者而言，重點不只防止模型出錯，也要防止工具、流程與部署被轉成攻擊面。","AI safety covers how models, tools, and deployment pipelines can be misused or fail in production: vulnerability discovery, malicious repositories, access controls, approval gates, and governance. For technical teams, the issue is not only model behavior but the attack surface around it.",[],15]