Given two measure spaces $(X,\mathcal{A},\mu)$ and $(Y,\mathcal{B},\nu)$, we can define the product measure $\mu\times\nu$ on the algebra $\mathcal{A}\times\mathcal{B}$ defined as the $\sigma$-algebra generated by all measurable rectangles in $X\times Y$.
Assuming $\mu$ and $\nu$ are complete, $\mu\times\nu$ need not be. For example, consider $A\in\mathcal{A}$ with $A\ne \emptyset$ but $\mu(A)=0$, and $B\in\mathcal{P}(Y)-\mathcal{B}$ (if $\mathcal{P}(Y)-\mathcal{B}\ne\emptyset$). Then $A\times B\notin \mathcal{A}\times\mathcal{B}$, but $A\times Y\supset A\times B$ is a null set.
How can we prove that $A\times B\notin \mathcal{A}\times\mathcal{B}$? It is not obvious to me that we cannot generate $A\times B$ from the measurable rectangles. I cannot think of a way to have $A\times B\in \mathcal{A}\times\mathcal{B}$, however I do not know how I would prove that it is impossible?