<p>Audio manipulation technologies continue to advance, evolving from traditional voice conversion and text-to-speech systems towards highly realistic spoofed audio. They have become a growing security concern, underscoring the urgent need for robust and intelligent detection approaches. Literature shows that classical models have performed well in detecting manipulated audio. However, these models are often resource-intensive due to a large number of trainable parameters. To address this, we introduce a novel hybrid classical-quantum model, WavLM-QLSTM, which leverages a quantum-inspired Long Short-Term Memory (QLSTM) network for efficient temporal modelling. The primary goal of the research is to reduce the large trainable parameters of classical LSTM with a highly efficient QLSTM, which drastically reduces model complexity while preserving performance. The proposed model was trained on the ASVspoof 2019 Logical Access (LA) spoofing dataset and evaluated against its classical counterpart on both the ASVspoof 2019 LA evaluation set and the challenging, unseen attacks of the ASVspoof 2021 Deepfake (DF) set. Our results show that the QLSTM component achieves a 99.5% parameter reduction compared to a classical LSTM (5,396 vs. 1.05M parameters). The hybrid quantum model obtains a competitive Equal Error Rate (EER) of 1.44% on the ASVspoof 2019 evaluation set. On the challenging, out-of-distribution ASVspoof 2021 Deepfake (DF) set, it achieves an EER of 11.38%. However, this is coupled with a significant performance trade-off, highlighting key limitations in its current generalisation capability. The work demonstrates the viability of using quantum-inspired components to build more efficient and robust spoofed audio detection systems.</p>