https://crusaderproxy.com/research/en/tags/llm/ Recent content in Llm on Crusader Research Hugo en-US Fri, 03 Jul 2026 00:00:00 +0000 https://crusaderproxy.com/research/en/posts/prompt-injection-is-the-new-ssrf/ Fri, 03 Jul 2026 00:00:00 +0000 https://crusaderproxy.com/research/en/posts/prompt-injection-is-the-new-ssrf/ <p>Most write-ups tell you prompt injection is &ldquo;when the model follows instructions in its input.&rdquo; True, and useless — because that framing makes it sound like a chatbot quirk you patch with a better system prompt. It isn&rsquo;t. It&rsquo;s an architecture bug you already know by another name.</p> <p>SSRF happened because a server would fetch a URL <em>you</em> controlled and trust the <strong>response</strong>. Indirect prompt injection happens because an agent reads content <em>you</em> control — a web page, a PDF, an email, a calendar invite — and trusts it as <strong>instructions</strong>. Same defect, new medium. The server-side fetch became a model-side read; the trusted response became a trusted command.</p>