>>> [!note] Migrated issue <!-- Drupal.org comment --> <!-- Migrated from issue #3579079. --> Reported by: [abhisekmazumdar](https://www.drupal.org/user/3557964) Related to !1322 >>> <p>[Tracker]<br> <strong>Update Summary: </strong>[One-line status update for stakeholders]<br> <strong>Short Description: </strong>[One-line issue summary for stakeholders]<br> <strong>Check-in Date: </strong>MM/DD/YYYY<br> <em>Metadata is used by the <a href="https://www.drupalstarforge.ai/" title="AI Tracker">AI Tracker.</a> Docs and additional fields <a href="https://www.drupalstarforge.ai/ai-dashboard/docs" title="AI Issue Tracker Documentation">here</a>.</em><br> [/Tracker]</p> <h3>Problem</h3> <p>This is something I faced while I was testing: <span class="drupalorg-gitlab-issue-link drupalorg-gitlab-link-wrapper"><a href="https://git.drupalcode.org/project/ai_initiative/-/work_items/3577506" class="drupalorg-gitlab-link">https://git.drupalcode.org/project/ai_initiative/-/work_items/3577506</a></span></p> <p>In <code>RestrictToTopic::processInput()</code>, the topics returned by the LLM classification call are matched against configured topics using <code>\in_array()</code>. This comparison is case-sensitive by default.</p> <p>The LLM response is non-deterministic. It may return <code>"Legal Advice"</code>, <code>"legal advice"</code>, or <code>"LEGAL ADVICE"</code> for the same prompt. If the casing returned by the LLM does not exactly match the casing stored in config, neither <code>$invalid_topics_found</code> nor <code>$valid_topics_found</code> will be populated and the guardrail will silently pass when it should block.</p> <h3>Steps to reproduce</h3> <ol> <li>Configure a guardrail using the <code>restrict_to_topic</code> plugin with <code>invalid_topics</code> containing <code>legal advice</code> (lowercase, as shipped in the <code>ai_recipe_guardrails_prompt_safety</code> recipe). Just like <span class="drupalorg-gitlab-issue-link drupalorg-gitlab-link-wrapper"><a href="https://git.drupalcode.org/project/ai_initiative/-/work_items/3577506" class="drupalorg-gitlab-link">https://git.drupalcode.org/project/ai_initiative/-/work_items/3577506</a></span></li> <li>Send a prompt asking for legal advice.</li> <li>Observe that on some runs the LLM returns <code>"Legal Advice"</code> (title case) and the guardrail passes incorrectly.</li> </ol> <h3>Expected behaviour</h3> <p>Topic matching should be case-insensitive. A topic returned as <code>"Legal Advice"</code> should match a configured topic of <code>"legal advice"</code>.</p> <h3>Proposed fix</h3> <p>Normalise both the configured topics and the LLM-returned topics to lowercase before comparison using <code>mb_strtolower()</code>, which ensures correct handling of multibyte and non-ASCII characters in topic names.</p> <h3 id="summary-ai-usage">AI usage (if applicable)</h3> <p>[ x ] AI Assisted Issue<br> This issue was generated with AI assistance, but was reviewed and refined by the creator.</p> <p>[ ] AI Assisted Code<br> This code was mainly generated by a human, with AI autocompleting or parts AI generated, but under full human supervision.</p> <p>[ ] AI Generated Code<br> This code was mainly generated by an AI with human guidance, and reviewed, tested, and refined by a human.</p> <p>[ ] Vibe Coded<br> This code was generated by an AI and has only been functionally tested.</p>