<?xml version="1.0" encoding="utf-8"?>
<feed xmlns="http://www.w3.org/2005/Atom">
  <generator uri="https://jekyllrb.com/" version="4.4.1">Jekyll</generator>
  <link href="https://adamdaniel.ai/tags/quotes/feed.xml" rel="self" type="application/atom+xml" />
  <link href="https://adamdaniel.ai/tags/quotes/" rel="alternate" type="text/html" />
  <updated>2026-05-28T19:35:00+00:00</updated>
  <id>https://adamdaniel.ai/tags/quotes/feed.xml</id>
  <title type="html">Adam Daniel: quotes</title>
  <subtitle>Posts tagged &quot;quotes&quot; on Adam Daniel.</subtitle>
  <author>
    <name>Adam Daniel</name>
    <email>adam@adamdaniel.ai</email>
  </author>
  <entry>
    <title type="html">Quoting Anthropic: Opus 4.8 Safety “somewhat less robust”</title>
    <link href="https://adamdaniel.ai/blog/quoting-anthropic-opus-4-8-safety-somewhat-less-robust/" rel="alternate" type="text/html" title="Quoting Anthropic: Opus 4.8 Safety “somewhat less robust”" />
    <published>2026-05-28T19:35:00+00:00</published>
    <updated>2026-05-28T19:35:00+00:00</updated>
    <id>https://adamdaniel.ai/blog/quoting-anthropic-opus-4-8-safety-%E2%80%9Csomewhat-less-robust%E2%80%9D</id>
    <content type="html" xml:base="https://adamdaniel.ai/blog/quoting-anthropic-opus-4-8-safety-somewhat-less-robust/">&lt;blockquote&gt;
  &lt;p&gt;&lt;strong&gt;Agentic safety.&lt;/strong&gt; Although it shows improvements in some areas (such as refusing malicious requests), we found Opus 4.8 to be somewhat less robust than Opus 4.7 in several agentic contexts (such as vulnerability to prompt injection attacks). However, the application of our safeguards closes the gap between the models in practice. […]&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;- Anthropic, &lt;a href=&quot;https://www.anthropic.com/claude-opus-4-8-system-card&quot;&gt;System Card: Claude Opus 4.8&lt;/a&gt;&lt;/p&gt;</content>
    <author>
      <name>Adam Daniel</name>
      <email>adam@adamdaniel.ai</email>
    </author>
    <category term="quotes" />
    <summary type="html">Agentic safety. Although it shows improvements in some areas (such as refusing malicious requests), we found Opus 4.8 to be somewhat less robust than Opus 4.7 in several agentic contexts (such as vulnerability to prompt injection attacks). However, the application of our safeguards closes the gap between the models in practice. […]</summary>
  </entry>
</feed>
