跳转至

04 部署者義務與治理實務 (Deployer Obligations & Governance)

🇬🇧 Switch to English Section 🇭🇰 切換至中文區塊

法定核心依據 / Legal Basis: Regulation (EU) 2024/1689 Article 4, Article 14, Article 26 & Article 27

🇭🇰 部署者義務解析 (中文版)

4.1 全員 AI 素養義務 (Article 4 - 2025 年 2 月生效)

所有組織必須確保負責操作、管理或監督 AI 的員工具備足夠的 AI 素養（AI Literacy）。

4.2 高風險 AI 部署者 6 大法定核心義務 (Article 26)

  1. 依指引操作：嚴格遵循 Provider 的使用說明書操作。
  2. 輸入數據控制：確保輸入數據具高度相關性與代表性。
  3. 自動日誌保存（Logs）：系統自動日誌必須妥善保存至少 6 個月。
  4. 有效人工監督 (Human Oversight)：指派專人監督，具備手動覆寫（Override）與一鍵終止權限。
  5. 異常停用與通報：發現重大風險立即暫停使用並通報當局。
  6. 員工與工會告知：在職場啟用高風險 AI 前，必須書面通知受影響員工與代表。

4.3 基本權利影響評估 (FRIA - Article 27)

公法機構及在信貸審批或人身保險定價中部署高風險 AI 的私人機構，上線前必須完成 FRIA 評估。

🇬🇧 Deployer Obligations & Governance (English Version)

4.1 Universal AI Literacy (Article 4 - Effective Feb 2025)

Organizations must ensure that personnel operating, managing, or overseeing AI systems maintain verified AI Literacy.

4.2 Six Core High-Risk Deployer Obligations (Article 26)

  1. Instructions of Use: Strict compliance with vendor manuals.
  2. Input Data Quality: Ensure data relevance and representativeness.
  3. Log Retention: Maintain automated logs for at least 6 months.
  4. Meaningful Human Oversight: Assign competent reviewers with override power.
  5. Incident Escalation: Suspend operations and notify authorities upon severe malfunctions.
  6. Worker Notification: Prior written notice to workers and unions before workplace deployment.

4.3 Fundamental Rights Impact Assessment (FRIA - Article 27)

Mandatory for public bodies and private entities deploying AI for credit scoring or life/health insurance pricing.