Suno aims to legitimize AI music with watermarks

Suno, one of the most prolific sources of AI-generated music, announced in a blog post that it will add watermarks to all audio outputs from its models and adopt policy changes to align with emerging industry standards for AI content labeling. CEO Mikey Shulman stated that this is necessary to meet those standards, and the watermarks will be invisible codes embedded in the waveform, detectable by trained systems, allowing partners to combat fraud and misuse. Shulman did not specify whether Suno will use an in-house solution or a ready-made one like Google's SynthID, which Google recently began licensing and claims has labeled 60,000 years of audio from its Gemini models and over 100 billion images and videos. The technology is said to be durable, tamper-resistant, and quality-neutral, but no system is perfect. If watermarks are broken, all previously generated tracks could lose their AI labels. Moreover, other platforms and open models could still generate unlabeled AI music, highlighting a broader challenge for watermarking AI content.