Unverified50% confidenceFactExact time
The two record labels allege that Suno's v6 model was trained on user-generated outputs from its previous models, which were themselves trained on music data scraped without authorization from platforms such as YouTube.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/26/2026
First Seen
Valid until: 12/25/2026
Sources
Related Entities
Related Claims
UnverifiedThe Suno v6 model is trained on licensed music rather than the music data used to train previous versions of the AI.78% similarUnverifiedSuno officially states that V6's training used licensed data from partners including Warner Music.74% similarUnverifiedPP-OCRv6采用了更轻量化的骨干网络与知识蒸馏训练策略，在保持高精度的同时压缩了模型体积56% similarUnverifiedv6 was trained from scratch using an entirely new dataset that does not contain the same data used by previous models.55% similarUnverifiedThe training pipeline involves first training an image baseline model, using the image model to warm-start an eight-frame video baseline model, and ultimately generating 5.6-second inference outputs.55% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/951711API
curl https://kongchang.com/api/v1/knowledge/claims/951711MCP
get_claim(id=951711)