No, an arbitrary Zariski dense subset need not contain any non-empty Zariski open subset. Take $S = \mathbb{Z} = \{ \dots, -2, -1, 0, 1, 2, \dots \} \subset \mathbb{A}^1_{\mathbb{C}}$. Since the non-trivial Zariski closed subsets of $\mathbb{A}^1_{\mathbb{C}}$ are the finite sets, it is easy to check that the Zariski closure of $S$ is $\mathbb{A}^1_{\mathbb{C}}$ and that the only open set $U \subset S$ is $U = \emptyset$.
Edit. You probably want to prove the following stronger statement: If $U$ is Zariski dense in $\mathbb{A}^n_K$ and $f$ is any non-constant polynomial, then $D(f) \cap U$ is Zariski dense in $\mathbb{A}^n_K$. Indeed, suppose for contradiction that it were not. Then $D(f) \cap U \subseteq Z(g)$ for some polynomial $g$ (where $Z(g)$ is the zero locus of $g$). But then $$U = (D(f) \cap U) \cup (U \setminus D(f)) \subseteq Z(g) \cup Z(f) = Z(fg),$$
contradicting the assumed density of $U$.