@spit_evil_olive_tips One example is the EU AI Act. Their requeriments for open source models are very lenient, only requiring summary of training data and disclosure of training details like the computational power used. But proprietary models, on the other hand, are required to implement content filtering etc.