How would you improve EU's AI Act?

I support the banning of social scoring and the strict regulations on facial recognition and a number of other things. However, I disagree with classifying foundation models as high risk, it seems influenced by the doomer crowd. Also, while transparency about data used in training is generally positive, it raises concerns about potential liability under current IP laws. But I guess that's more of a problem around copyright law that should be changed

just as patents drove the film industry to California. "AI" Acts will drive the AI industry to outside of EU.