<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>AI安全 on 1630.top 技术博客</title><link>https://www.1630.top/tags/ai%E5%AE%89%E5%85%A8.html</link><description>Recent content in AI安全 on 1630.top 技术博客</description><generator>Hugo</generator><language>zh-CN</language><lastBuildDate>Sat, 25 Jul 2026 08:00:00 +0800</lastBuildDate><atom:link href="https://www.1630.top/tags/ai%E5%AE%89%E5%85%A8/rss.xml" rel="self" type="application/rss+xml"/><item><title>AI Agent安全防护体系：Prompt注入检测、输出审核与访问控制完整实战</title><link>https://www.1630.top/articles/ai-agent-security-defense-guide.html</link><pubDate>Sat, 25 Jul 2026 08:00:00 +0800</pubDate><guid>https://www.1630.top/articles/ai-agent-security-defense-guide.html</guid><description>从实战角度搭建AI Agent三层安全防护体系：输入层Prompt注入检测（规则引擎+LLM二次检测）、执行层工具权限分级与代码沙箱、输出层敏感信息过滤，附完整Python代码。</description></item><item><title>你的AI系统真的安全吗？大模型红队测试自动化实战指南</title><link>https://www.1630.top/articles/llm-red-team-testing-automation-guide.html</link><pubDate>Wed, 22 Jul 2026 08:00:00 +0800</pubDate><guid>https://www.1630.top/articles/llm-red-team-testing-automation-guide.html</guid><description>2026年企业AI安全必修课。本文从四大攻击向量出发，用Python构建自动化红队测试流水线，包含Prompt注入测试、越狱检测、数据泄露测试框架和报告生成代码，并演示如何接入CI/CD持续守护AI系统安全。</description></item><item><title>大模型API工程化实战：限流、缓存、降级与成本优化完整指南</title><link>https://www.1630.top/articles/llm-api-engineering-production-guide.html</link><pubDate>Mon, 13 Jul 2026 08:00:00 +0800</pubDate><guid>https://www.1630.top/articles/llm-api-engineering-production-guide.html</guid><description>从单体API调用到企业级LLM网关，本文系统讲解大模型API工程化的六大核心模块：限流策略、语义缓存、故障降级、成本控制、重试机制与可观测性，附完整Python代码示例。</description></item></channel></rss>