Existing weakly supervised salient object detection (WSSOD) methods based on SAM and SAM2 still rely on prompts during training, necessitating the inclusion of additional prompts such as points or bounding boxes. This reliance contradicts the core objective of the SOD task, which aims to directly detect the most salient objects without such explicit guidance. To address these issues, we propose a Laplacian pyramid-inspired network built upon the fine-tuning of SAM2, termed SAM2-LPNet. Our SAM2-LPNet leverages a focal adapter to replace traditional prompts, enabling the SAM2 encoder to focus on the most salient objects in the image. Additionally, we reformulate the SAM2 decoder based on the Laplacian pyramid approach to reduce the impact of prompt absence on decoder performance. Extensive experiments on five mainstream SOD benchmarks demonstrate that our method surpasses state-of-the-art weakly and even fully supervised approaches.