<?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>Prompt-Injection on Local First AI</title><link>https://localfirstai.eu/tags/prompt-injection/</link><description>Recent content in Prompt-Injection on Local First AI</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Wed, 22 Jul 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://localfirstai.eu/tags/prompt-injection/index.xml" rel="self" type="application/rss+xml"/><item><title>We Red-Teamed Our Own Bot</title><link>https://localfirstai.eu/posts/2026-07-22-we-red-teamed-our-own-bot/</link><pubDate>Wed, 22 Jul 2026 00:00:00 +0000</pubDate><guid>https://localfirstai.eu/posts/2026-07-22-we-red-teamed-our-own-bot/</guid><description>The CasaSol Guide bot resisted every prompt injection and session-extraction attack we threw at it. Then we found the vulnerability that wasn&amp;#39;t in the model at all — it was in who gets to decide what the model believes.</description></item></channel></rss>