Let $\kappa$ be a cardinal, and let $P$ be a poset. Let $\mathcal{P}_\kappa(P)$ denote the poset of $\kappa$-small subposets of $P$ and let $\mathcal{P}_\kappa^\downarrow(P)\subseteq\mathcal{P}_\kappa(P)$ be the subposet consisting of those subposets that are downward-closed. Then according to a reliable source, when $\kappa$ is regular, we can show that $\mathcal{P}^\downarrow_\kappa(P)$ is $\kappa$-filtered because given some $\kappa$-small family of $\kappa$-small subposets, $$A_i:I\to \mathcal{P}_\kappa(P)\quad |I|<\kappa$$

the downward closure of the union over this family, $\operatorname{Cl}^\downarrow(\bigcup_{i\in I}A_i)$, is $\kappa$-small (which gives a majorant for the family $A_i$).

However, since I have no experience at all working with regular cardinals, I'm not really sure how to make heads or tails of this. Why does the regularity of $\kappa$ imply that the downward-closure of that union is $\kappa$-small?