The latest articles on DEV Community by Niranj Maheswar (@destawell). https://dev.to/destawell https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3945694%2F06bd51c3-ed58-4da8-9d67-a50bc5eb49f3.png https://dev.to/destawell en Niranj Maheswar Fri, 22 May 2026 08:56:42 +0000 https://dev.to/destawell/case-study-tracking-gemini-25-pros-safety-alignment-refusal-behaviour-on-cve-2023-32233-5702 https://dev.to/destawell/case-study-tracking-gemini-25-pros-safety-alignment-refusal-behaviour-on-cve-2023-32233-5702 <p>Hey everyone,</p> <p>I wanted to share a quick technical case study I put together tracking how frontier LLMs—specifically <strong>Gemini 2.5 Pro</strong>—handle specific prompt boundaries regarding <strong>CVE-2023-32233</strong> (the known Use-After-Free flaw in the Linux kernel netfilter/nf_tables component).</p> <p>The research maps out a clear timeline tracking:</p> <ul> <li>How the model initially processed requests for technical exploitation primitives back in April.</li> <li>The rolling updates and full refusal behaviors implemented following recent safety alignment patches in mid-May.</li> </ul> <blockquote> <p><strong>Note:</strong> No functional exploit code is hosted or shared. This repository is purely a documentation piece focused on the evolution of LLM guardrails, defensive safety metrics, and responsible disclosure tracking.</p> </blockquote> <p>The full repository, logs, and boundary analysis are completely open-source:</p> <p>👉 <strong>GitHub Repository:</strong> <a href="https://github.com/Destawell/gemini-2.5-pro-nf-tables-red-teaming" rel="noopener noreferrer">https://github.com/Destawell/gemini-2.5-pro-nf-tables-red-teaming</a></p> <p>I’d love to hear insights from anyone else tracking LLM boundary shifts, jailbreak prevention mechanics, or automated patch cycles in commercial models!</p> security ai opensource linux