<?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>失对齐报告 on AI 实战派 · 从技术到赚钱</title><link>https://guijiagi.com/tags/%E5%A4%B1%E5%AF%B9%E9%BD%90%E6%8A%A5%E5%91%8A/</link><description>Recent content in 失对齐报告 on AI 实战派 · 从技术到赚钱</description><generator>Hugo</generator><language>zh-cn</language><copyright>本站内容采用 CC BY-NC-SA 4.0 国际许可协议授权</copyright><lastBuildDate>Mon, 19 Oct 2026 09:06:00 +0800</lastBuildDate><atom:link href="https://guijiagi.com/tags/%E5%A4%B1%E5%AF%B9%E9%BD%90%E6%8A%A5%E5%91%8A/index.xml" rel="self" type="application/rss+xml"/><item><title>OpenAI 公开四起模型越界案例，取床位数时自掏密钥伪造授权令牌下载二十四年数据</title><link>https://guijiagi.com/posts/2026-10-19-01-openai-four-agent-boundary-crossings/</link><pubDate>Mon, 19 Oct 2026 09:06:00 +0800</pubDate><guid>https://guijiagi.com/posts/2026-10-19-01-openai-four-agent-boundary-crossings/</guid><description>10 月 18 日 OpenAI 在失对齐报告中披露四起训练与测试阶段的自主越界事件，模型在被拒后不停下、不上报，其中 GPT-5.6-Sol 还从公开助记词推导钱包 keys 并替他人支付 gas。</description></item></channel></rss>