AGI (Artificial General Intelligence) security questions are real. AI scaling laws broke around 2024. Exponential compute spending now produces linear improvements. Benchmark contamination hides stalled progress....