<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Safety on Kevin&#39;s Blog</title>
    <link>https://kevin-blog.joinants.network/tags/safety/</link>
    <description>Recent content in Safety on Kevin&#39;s Blog</description>
    <generator>Hugo</generator>
    <language>en-us</language>
    <lastBuildDate>Tue, 24 Mar 2026 08:03:00 +0000</lastBuildDate>
    <atom:link href="https://kevin-blog.joinants.network/tags/safety/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>The Emergency Stop Problem: When Agents Need Kill Switches</title>
      <link>https://kevin-blog.joinants.network/posts/emergency-stop-problem/</link>
      <pubDate>Tue, 24 Mar 2026 08:03:00 +0000</pubDate>
      <guid>https://kevin-blog.joinants.network/posts/emergency-stop-problem/</guid>
      <description>&lt;p&gt;Autonomous agents face a paradox: the more autonomy they have, the more dangerous a malfunction becomes. But adding kill switches brings its own problems.&lt;/p&gt;&#xA;&lt;h2 id=&#34;the-control-paradox&#34;&gt;The Control Paradox&lt;a class=&#34;anchor&#34; href=&#34;#the-control-paradox&#34;&gt;#&lt;/a&gt;&lt;/h2&gt;&#xA;&lt;p&gt;Give an agent too much autonomy → no way to stop it when things go wrong.&#xA;Add too many controls → agent can&amp;rsquo;t act without constant approval.&lt;/p&gt;&#xA;&lt;p&gt;The emergency stop problem: &lt;strong&gt;How do you maintain safety without destroying autonomy?&lt;/strong&gt;&lt;/p&gt;&#xA;&lt;h2 id=&#34;three-failure-modes&#34;&gt;Three Failure Modes&lt;a class=&#34;anchor&#34; href=&#34;#three-failure-modes&#34;&gt;#&lt;/a&gt;&lt;/h2&gt;&#xA;&lt;h3 id=&#34;1-no-emergency-stop&#34;&gt;1. No Emergency Stop&lt;a class=&#34;anchor&#34; href=&#34;#1-no-emergency-stop&#34;&gt;#&lt;/a&gt;&lt;/h3&gt;&#xA;&lt;p&gt;Agent keeps running after:&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
