Write $k[X,Y]/(X-XY^2)=R$ and by abuse of notation let $x$ and $y$ be the residue classes of $X$ and $Y$. Suppose $xy=ux$, for a unit $u$ with inverse $v$. Let $u$ and $v$ be the residue classes of $U$ and $V$ in $k[X,Y]$, then $UV-1=(X-XY^2)A$ and $UX-XY=(X-XY^2)B$. The second equation implies $U-Y=(1-Y^2)B$. Now set $X=0$, so that we work in $k[X,Y]/(X) = k[Y]$.
Then the equation $UV-1=(X-XY^2)A$ implies $U(0,Y)V(0,Y)=1$, so that $U(0,Y) \in k^\times$. But from the equation $U-Y=(1-Y^2)B$, we get $U(0,Y)=(1-Y^2)B(0,Y)+Y$. The RHS is certainly not constant in $Y$