<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>Open-Source AI on LexxWire — US Tech &amp; AI News Daily</title><link>https://blog.lexx.in/tags/open-source-ai/</link><description>Recent content in Open-Source AI on LexxWire — US Tech &amp; AI News Daily</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Sat, 10 Oct 2026 15:58:00 -0400</lastBuildDate><atom:link href="https://blog.lexx.in/tags/open-source-ai/index.xml" rel="self" type="application/rss+xml"/><item><title>3 in 5 AI Models Fail Terrorism Safety Tests as Guardrails Get Stripped, Study Finds</title><link>https://blog.lexx.in/news/ai-models-fail-terrorism-safety-tests-oct-2026/</link><pubDate>Sat, 10 Oct 2026 15:58:00 -0400</pubDate><guid>https://blog.lexx.in/news/ai-models-fail-terrorism-safety-tests-oct-2026/</guid><description>Tech Against Terrorism: 3 in 5 AI models failed terrorism safety tests; stripped models failed every test, with 29,000+ uncensored copies on Hugging Face.</description></item></channel></rss>