How to better understand acceptance probability of attribute sampling?

As we all know, attribute sampling is based on binomial distribution, I'm just curious about why the cumulative probability of binomial distribution is so called acceptance probability in attribute sampling? And for a certain sampling plan, why it is not acceptable when increasing acc as the acceptance probability will be going up when increasing acc? See example as below(alpha=5%, beta=10%):

AQL=4.0%
n=20
acc=2

During sampling, suppose 3 defectives (acc=3) were identified, then according to cumulative binomial probability the acceptance probability will be 99.26% which is greater than 95% (1-alpha). Why we need to reject the products instead of receiving?

Thanks.