<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>llm on Dark Bio Blog</title>
    <link>https://blog.dark.bio/tags/llm/</link>
    <description>Recent content in llm on Dark Bio Blog</description>
    <generator>Hugo</generator>
    <language>en-us</language>
    <copyright>Copyright © 2026, Dark Bio AG.</copyright>
    <lastBuildDate>Mon, 09 Feb 2026 00:00:00 +0000</lastBuildDate>
    <atom:link href="https://blog.dark.bio/tags/llm/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Leaky LLMs: Accident or Nature?</title>
      <link>https://blog.dark.bio/2026/02/09/leaky-llms-accident-or-nature/</link>
      <pubDate>Mon, 09 Feb 2026 00:00:00 +0000</pubDate>
      <guid>https://blog.dark.bio/2026/02/09/leaky-llms-accident-or-nature/</guid>
      <description>&lt;p&gt;Approximately two weeks ago, OpenClaw (originally Clawdbot renamed MoltBot)&#xA;became an overnight sensation. It allowed anyone to run their own proactive&#xA;AI agent, which could just &lt;em&gt;&lt;strong&gt;do&lt;/strong&gt;&lt;/em&gt; stuff without having to be prompted. This had&#xA;a huge appeal in itself, but was even further exacerbated by the appearance&#xA;of moltbook, a kind of Reddit/Facebook mix for AI agents. People went wild,&#xA;a lot of money was made, a lot of agents were launched. On the one hand, it&#xA;is an amazing and fun development. On the other, it&amp;rsquo;s a catastrophe waiting&#xA;to happen. Problem is, people don&amp;rsquo;t understand the tech yet, but have a big&#xA;fear of missing out on it.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
