https://dl4sec.com/tags/llm/ Recent content in LLM on DL4Sec Hugo -- 0.147.7 en-us Fri, 07 Aug 2026 21:27:38 +0200 https://dl4sec.com/blog/generative-ai/hack-the-bus/ Fri, 07 Aug 2026 21:27:38 +0200 https://dl4sec.com/blog/generative-ai/hack-the-bus/ Post-training safety looks robust until you find the one direction in activation space that carries refusal. Ablate it and the model stops saying no, which tells us alignment is thinner than it appears.