Let $\mu$ a finite, atomless (non-atomic) Borel measure on $\mathbb{R}^n$. Is it necessarily true that $\mu\big(\partial B_{\epsilon}(x)\big)=0$ for any open ball $B_{\epsilon}(x)$?
This is true for the Lebesgue measure but most standard arguments use translation-invariance. I also know this holds for Borel product measures with atomless marginals: is it true in general for any finite, atomless Borel measure?