Office AI 安全说明 / Office AI Security Guide

结论 / Conclusion

安全使用的核心是可信来源、最小权限、可回退文件和人工复核。只使用当前公开版本和经过审核的脚本方案，不执行来源不明的文件，不把密钥或真实业务数据提交到公开页面；发现异常时保留证据并通过支持或社区规范入口报告。

Safe use depends on trusted sources, least privilege, restorable files, and human review. Use only the current public release and reviewed script solutions, do not run files from unknown sources, and never submit keys or real business data to public pages. Preserve evidence and report anomalies through the support or community-guideline paths.

安全基线 / Security baseline

• 下载前核对版本、来源、HTTPS 地址和校验信息 / Verify version, source, HTTPS URL, and checksum before downloading.
• 重要文件先备份，在副本和小样本上验证自动化结果 / Back up important files and validate automation on copies and small samples.
• 脚本执行前检查来源、权限、读写范围、网络依赖和回退方式 / Before running a script, check provenance, permissions, read/write scope, network dependencies, and rollback.
• 不发布 token、密钥、真实业务文件、未脱敏日志、恶意脚本或无关内容 / Do not publish tokens, keys, real business files, unredacted logs, malicious scripts, or unrelated content.
• 对 AI 输出、公式、脚本和自动化结果执行人工验收 / Perform human acceptance checks on AI output, formulas, scripts, and automation results.

发现问题时 / When you find a problem

1. 停止继续覆盖原文件，保留可复现副本 / Stop overwriting the original and keep a reproducible copy.
2. 记录版本、宿主应用、Windows 版本、时间、错误和最短步骤 / Record version, host application, Windows version, time, error, and shortest steps.
3. 脱敏后通过支持中心或合适的社区分类提交 / Redact sensitive data and submit through support or the appropriate community category.
4. 不要把安全事件细节公开到普通主题或第三方群聊 / Do not publish security-incident details in ordinary topics or third-party chats.

官方入口 / Official paths:

• 安全页面 / Security page: /security
• 隐私说明 / Privacy: /docs/privacy.md
• 故障排查 / Troubleshooting: /docs/troubleshooting.md
• 支持中心 / Support: /support