1M original samples: How Pozalabs creates ethical AI music

South Korean startup Pozalabs has publicly defended its training methods, revealing it uses only in-house created content for its AI models.

Responding to inquiries from French collecting society SACEM, Pozalabs CEO Wongil Huh confirmed the company’s commitment to ethical AI development.

---

💡3 key takeways:

• Pozalabs creates AI music using only in-house composed training data
• 15 composers generated over 1 million samples for ethical AI training
• Company proves ethical AI music development possible without unauthorized content

---

Building an ethical dataset

Their model relies on over 1 million MIDI sound samples and 50,000 vocal samples, all created by their 15 professional in-house composers, avoiding any unauthorized use of copyrighted material.

This stance places Pozalabs in direct contrast with competitors who scrape vast amounts of commercial music without permission like Suno and Udio. While this ethical approach may limit their training data compared to companies using unauthorized content, it aligns with growing industry demands for responsible AI development, similar to recent initiatives like AIOK’s mission to develop industry standards.

Industry impact and future outlook

The company’s platforms, LAIVE and eapy, enable users to create original music, lyrics, and vocals using their proprietary AI algorithms. This approach mirrors other ethical initiatives in the industry, such as Fairly Trained’s certification program for ethical AI development.

This development comes at a crucial time when major music organizations are increasingly scrutinizing AI companies’ training practices. As explored in our coverage of HarmonyCloak’s protection against AI copying, the industry is actively seeking solutions to ensure ethical AI use in music production.

Pozalabs demonstrates that ethical AI development in music is achievable through careful attention to training data sources. While other companies rely on unauthorized datasets, their approach of using in-house created content shows a practical path forward for responsible AI development in the music industry.