Đã lưu trong:
| Tác giả chính: | |
|---|---|
| Định dạng: | Recurso digital |
| Ngôn ngữ: | |
| Được phát hành: |
Zenodo
2025
|
| Những chủ đề: | |
| Truy cập trực tuyến: | https://doi.org/10.5281/zenodo.18004672 |
| Các nhãn: |
Thêm thẻ
Không có thẻ, Là người đầu tiên thẻ bản ghi này!
|
Mục lục:
- <p>High-risk artificial intelligence systems increasingly operate at execution speeds that exceed human or procedural oversight. While the European Union Artificial Intelligence Act (EU AI Act) emphasizes ex-ante risk mitigation, transparency, and accountability, it remains largely implementation-neutral regarding how unlawful or unsafe actions should be technically prevented at the moment of execution.</p> <p>This paper presents <strong>Action-Specific Oversight (ASiSO)</strong> as a conceptual governance architecture addressing this execution-level control gap. ASiSO introduces an external, neutral veto layer operating immediately prior to action execution, enabling deterministic interruption or blocking of prohibited actions before harm occurs.</p> <p>The paper positions ASiSO within regulatory and insurance contexts and provides a structured, non-normative mapping between key EU AI Act requirements and ASiSO’s functional contributions. ASiSO is presented as a governance-enabling architecture rather than a product, standard, or certification.</p>