<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0">
  <channel>
    <title>Mizara Blog</title>
    <link>https://mizara.ai/blog</link>
    <description>Engineering notes on authorizing AI agents.</description>
    <language>en-us</language>
    
    <item>
      <title>AI Agents Need More Than Authorization</title>
      <link>https://mizara.ai/blog/agents-dont-have-an-authorization-problem</link>
      <guid>https://mizara.ai/blog/agents-dont-have-an-authorization-problem</guid>
      <pubDate>Fri, 10 Jul 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[While prototyping a support agent, we hit a well-formed request that never should have gone through - no prompt injection, no bug. Notes on why authorization doesn't cover this, and what does.]]></description>
    </item>
  </channel>
</rss>