The rapid development of technology has led to an increase in the number of devices that rely on position, velocity, and time (PVT) information to perform their functions. As such, the Global Navigation Satellite Systems (GNSS) have been adopted as one of the most promising solutions to provide PVT. Consequently, there are renewed efforts aimed at enhancing GNSS capabilities to meet emerging use cases and their requirements. For example, GNSS is evolving to rely on low-earth-orbit satellites, shifting the focus from traditional medium-earth-orbit satellites. Unfortunately, these developments also bring forth higher risks of interference signals such as spoofers, which pose serious security threats. To address this challenge, artificial intelligence (AI)-inspired solutions are being developed to overcome the limitations of conventional mathematics-based approaches, which have proven inflexible when dealing with diverse forms of interference. In this paper, we advance this direction by proposing a meta-learning framework that enables GNSS receivers to detect various types of spoofers. Specifically, our approach exploits the radio frequency fingerprints present in the signal at both the pre-correlation and post-correlation stages of the receiver. The proposed solution has superior generalization properties compared to the state-of-the-art solutions. Numerical results demonstrate that our proposed solution significantly detects spoofers of different forms, with spoofing detection accuracies of more than 95% on multiple datasets from the Texas Spoofing Test Battery (TEXBAT) and the Oak Ridge Spoofing and Interference Test Battery (OAKBAT) repositories