Summarize the key differences between the EU AI Act and the proposed US AI safety guidelines regarding "High-Risk" foundational models. Analyze the potential compliance burden for mid-sized software companies and suggest a proactive risk-mitigation framework.