The fifth edition of the ASVspoof challenge advances research on speech spoofing and deepfake detection by introducing a large-scale, crowdsourced corpus. ASVspoof 5 contains data from roughly 2000 speakers and 32 attack algorithms—including adversarial, text-to-speech, and voice-conversion systems—collected under diverse acoustic conditions. The paper details the corpus design, partitioning protocols, and baseline validations for both automatic speaker verification and spoof detection systems, establishing a foundation for future robustness research in speech forensics.