Ask HN: How AI teams source and license training data?

I'm deep in research on how AI teams actually source and license training data (text, audio, video, synthetic). Not the theory, but real, messy, day-to-day process. I'm NOT pitching or selling anything. I'm having short 15-minute conversations with people who work on this daily, and the insights have been genuinely eye-opening. Happy to share what I'm learning in return. If you know someone who fits any of these, I'd massively appreciate an intro or a tag in the comments. Possible targets: ML engineers or data leads at companies training or fine-tuning LLMs. Anyone responsible for sourcing or procuring training data. Teams building domain-specific AI models (healthcare, legal, finance, speech) People working on multilingual model training