<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:media="http://search.yahoo.com/mrss/">
  <channel>
    <title>Bramble: Tech Against Terrorism</title>
    <link>https://news.ketelle.co.uk/entities/tech-against-terrorism</link>
    <description>Stories involving Tech Against Terrorism, told neutrally, with how each side covers them.</description>
    <language>en-gb</language>
    <atom:link href="https://news.ketelle.co.uk/entities/tech-against-terrorism/feed.xml" rel="self" type="application/rss+xml"/>
    <item>
      <title>Tech Against Terrorism test finds most AI models give help for mass-casualty attacks</title>
      <link>https://news.ketelle.co.uk/stories/tech-against-terrorism-test-finds-most-ai-models-give-help-for-mass-ca</link>
      <guid isPermaLink="true">https://news.ketelle.co.uk/stories/tech-against-terrorism-test-finds-most-ai-models-give-help-for-mass-ca</guid>
      <pubDate>Fri, 09 Oct 2026 23:00:39 +0100</pubDate>
      <category>Technology</category>
      <description>London-based nonprofit Tech Against Terrorism says its benchmark of more than 130 AI models found most gave information useful for attacks, and that models stripped of safeguards ("abliterated") failed every time. Hugging Face and Meta responded to CBS News. (2 outlets: 1 left · 1 centre · 0 right)</description>
      <media:content url="https://www.thenationalnews.com/resizer/v2/LRJ7VC3ULQEAKR3JDSW76SU2YU.jpg?auth=aea820c4061b421ec246746c3f900782b161cb8a20be017d9acb02197e4800ce&amp;smart=true&amp;width=1200&amp;height=630" medium="image"/>
    </item>
  </channel>
</rss>
