<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0"><channel><title><![CDATA[Topics tagged with openweight]]></title><description><![CDATA[A list of topics that have been tagged with openweight]]></description><link>https://forum.fedi.dk/tags/openweight</link><generator>RSS for Node</generator><lastBuildDate>Thu, 29 Oct 2026 05:23:21 GMT</lastBuildDate><atom:link href="https://forum.fedi.dk/tags/openweight.rss" rel="self" type="application/rss+xml"/><pubDate>Invalid Date</pubDate><ttl>60</ttl><item><title><![CDATA[Backdooring #OpenWeight #AI Models]]></title><description><![CDATA[@thenewoil I did not know that you could do something like that, but it basically boils down to the same principles that you should hold on to when downloading executable software: Only download from trusted sources, and if possible verify the download with a published hash.And yes, I'm aware that you will never be able to have an absolute guarantee that there is nothing fishy hidden away somewhere, but getting your model directly from Ollama or a company profile (like Qwen or Google or Microsoft) on HuggingFace is a much safer bet than downloading an abliterated Qwen3-14B from user "xrdf" who only ever uploaded that one model...]]></description><link>https://forum.fedi.dk/topic/8c30b7f5-f66d-424b-8119-781562e0671d/backdooring-openweight-ai-models</link><guid isPermaLink="true">https://forum.fedi.dk/topic/8c30b7f5-f66d-424b-8119-781562e0671d/backdooring-openweight-ai-models</guid><dc:creator><![CDATA[madsenandersc@social.vivaldi.net]]></dc:creator><pubDate>Invalid Date</pubDate></item></channel></rss>