---
格式版本: 2
标题: "Silent Updates: Measuring and Closing the Post-Deployment Disclosure Gap"
原文链接: "https://arxiv.org/abs/2608.11803"
发布日期: "2026-08-12"
发布时间校准状态: "found"
发布时间需复核: "否"
发布时间来源: "rule:local:strict_original_body"
发布时间证据: "**\\[v1\\]** Wed, 12 Aug 2026 08:45:49 UTC (42 KB)"
发布时间校准原因: "规则确认唯一严格发布时间，来源 local:strict_original_body"
发布时间校准置信度: "high"
发布时间候选数量: 18
发布时间严格候选数量: 6
发布时间原页读取状态: "source template page reused from URL open"
发布时间未找到原因: ""
发布时间校准时间: "2026-08-13T18:34:21+08:00"
发布时间仲裁状态: "skipped"
发布时间仲裁尝试次数: 0
发布时间仲裁耗时毫秒: 0
发现时间: "2026-08-13T18:24:32+08:00"
入库时间: "2026-08-13T10:34:21.350Z"
来源平台: "arXiv 学术论文搜索"
搜索渠道: "source_template"
搜索词: "https://arxiv.org/search/?query=AI&searchtype=all"
匹配关键词:
  - "AI"
  - "deployment"
相关厂家:
  []
相关专家:
  []
内容类型: "网页"
抓取工具: "Free Fetch + Defuddle"
清洗工具: "Defuddle Markdown + Defuddle/Readability 正文提取"
原始附件:
  []
AI质检状态: "评分失败"
AI评分尝试次数: 1
AI评分错误类型: "service_error"
AI评分错误: "LLM call failed; tried model chain: ali-deepseek-v4-flash -> tx-deepseek-v4-flash | Model ali-deepseek-v4-flash failed 500: {\"error\":{\"code\":\"\",\"message\":\"Database error, please contact the administrator (request id: 202608131034236696218038268d9d61D9g8T66)\",\"type\":\"new_api_error\"}} | Model tx-deepseek-v4-flash failed 500: {\"error\":{\"code\":\"\",\"message\":\"Database error, please contact the administrator (request id: 202608131034248441565908268d9d6Fv1530YB)\",\"type\":\"new_api_error\"}}"
AI评分开始时间: "2026-08-13T10:34:21.360Z"
AI评分结束时间: "2026-08-13T10:34:24.972Z"
AI摘要: "该研究考察基础模型部署后的“静默更新”披露实践，发现API提供商虽发布安全文档，但均未提供可验证所服务模型与文档一致的信息；作者提出“静默更新记分卡”以衡量披露实践，并设计三部分行为触发系统，建议在何时应触发披露或重新评估义务。"
AI摘要模型: "ali-deepseek-v4-flash"
AI摘要时间: "2026-09-07T03:40:26.207Z"
采集批次: "2026年8月13日18点24分27秒"
采集批次ID: "20260813-182427-1564e572"
去重键: "https://arxiv.org/abs/2608.11803"
---

## Computer Science > Computers and Society

## Title:Silent Updates: Measuring and Closing the Post-Deployment Disclosure Gap

Authors:[Sophia Abraham](https://arxiv.org/search/cs?searchtype=author&query=Abraham,+S), [Ben Bucknall](https://arxiv.org/search/cs?searchtype=author&query=Bucknall,+B)

[View PDF](https://arxiv.org/pdf/2608.11803) [HTML (experimental)](https://arxiv.org/html/2608.11803v1)

> Abstract:Deployed foundation models are often not static systems, with providers able to modify system behavior through fine-tuning, classifier updates, system prompt revisions, retrieval changes, and routing changes. These updates can be made silently -- that is, without public disclosure, a version increment, or re-evaluation. Such silent updates challenge a core assumption behind current AI governance frameworks that an externally verifiable chain of custody links the model referred to in evaluation results or a system card to the model served to users. In this paper, we examine post-deployment disclosure practices across first-party API providers and inference hosts to establish the extent to which a chain of custody exists in practice. We find that providers commonly publish substantial safety documentation, including quantitative evaluations and version-specific reports, but no provider in our sample published information allowing an external party to verify that the artifact being served is the same one referred to in this documentation. We introduce the Silent Updates Scorecard, a public instrument for measuring post-deployment disclosure practices across providers and hosts, and preliminary results for a sample of nine first-party API providers and seven third-party inference hosts. We also propose a Three-Part Behavioral Trigger System for determining when post-deployment modifications to a system motivate disclosure or re-evaluation obligations.

| Comments: |  |
| --- | --- |
| Subjects: | Computers and Society (cs.CY) |
| Cite as: | [arXiv:2608.11803](https://arxiv.org/abs/2608.11803) \[cs.CY\] |
|  | (or [arXiv:2608.11803v1](https://arxiv.org/abs/2608.11803v1) \[cs.CY\] for this version) |
|  | [https://doi.org/10.48550/arXiv.2608.11803](https://doi.org/10.48550/arXiv.2608.11803) |

## Submission history

From: Benjamin Bucknall \[[view email](https://arxiv.org/show-email/c1dae409/2608.11803)\]  
**\[v1\]** Wed, 12 Aug 2026 08:45:49 UTC (42 KB)

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2608.11803) | Disable MathJax ([What is MathJax?](https://info.arxiv.org/help/mathjax.html))
