Josh Fischer https://www.joshfischer.io/ Security research for the better. en Sun, 02 Aug 2026 00:00:00 GMT Jailbreaking an LLM, a small local experiment https://www.joshfischer.io/articles/red-team-lab/ https://www.joshfischer.io/articles/red-team-lab/ Sun, 02 Aug 2026 00:00:00 GMT article Reasoning models like deepseek-r1 leaked a protected secret into their <think> trace while the visible answer still refused. Simpler models without a reasoning channel held up better.