https://cognaptus.com/tags/model-compliance/ Recent content in Model Compliance on Cognaptus Hugo -- 0.145.0 en-us Tue, 19 Aug 2025 00:00:00 +0000 https://cognaptus.com/blog/2025-08-19-forgetting-by-design-turning-gdpr-into-a-systems-problem-for-llms/ Tue, 19 Aug 2025 00:00:00 +0000 https://cognaptus.com/blog/2025-08-19-forgetting-by-design-turning-gdpr-into-a-systems-problem-for-llms/ <script> window.MathJax = { tex: { inlineMath: [['$', '$'], ['\(', '\)']], displayMath: [['$$', '$$'], ['\[', '\]']] } }; </script> <script async src="https://cdn.jsdelivr.net/npm/mathjax@3/es5/tex-mml-chtml.js"></script> <h2 id="tldr-for-operators">TL;DR for operators</h2> <p>A deletion request is not a prompt. It is not a “please forget” instruction, a fine-tuning vibe, or a compliance-flavoured model apology.</p> <p>The useful idea in <em>Unlearning at Scale: Implementing the Right to be Forgotten in Large Language Models</em> is much less mystical: make training reproducible enough that deletion can be executed like systems recovery.<sup id="fnref:1"><a href="#fn:1" class="footnote-ref" role="doc-noteref">1</a></sup> The paper treats training as a deterministic program, logs the minimal control inputs needed to replay that program, and then removes the requested data during replay. Under strict preconditions, the resulting parameters are bit-identical, in the training dtype, to the model that would have been produced if the forgotten examples had never been included.</p>