Systolic array-based architectures are one of the most used types of AI accelerators. Like all integrated circuits (ICs), systolic array-based AI accelerators are subjected to aging, which results in lower performance, and higher energy over time. In this work, we propose a simple aging attack that accelerates the degradation process. The attack exploits the negative bias temperature instability (NBTI) phenomenon in PMOS. The attack takes advantage of the commutative property of addition, and interchanges input connections at specific full adders. As we show later, this simple tampering increases the likelihood of negative bias and accelerates aging on PMOS without affecting correctness. Experimental results indicate that the proposed attack leads to an 11% increase in execution time after one month, 37% after one year, and 72% after four years compared to the initial delay. This degradation results in a 66% average accuracy loss after four years, demonstrating the effectiveness of this attack.