The following is a remark from my textbook:
One might think that $\sigma(\mathcal{G})$ (a $\sigma$-algebra generated by $\mathcal{G}$) can be explicitly constructed for any given $\mathcal{G}$ (where $\mathcal{G} \subset \mathcal{P}(X)$ is a system of sets and $\mathcal{P}(X)$ is the power set) by adding to the family $\mathcal{G}$ all possible countable unions of its members and their complements: $$\mathcal{G}_{\sigma c} = \bigg\{ \bigcup_{n \in N} G_n, \bigg(\bigcap_{n \in N} G_n\bigg)^c : G_n \in \mathcal{G} \bigg\}. $$ But $\mathcal{G}_{\sigma c}$ is not necessarily a $\sigma$-algebra.
My thinking is if I let $\mathcal{G} = \{A,B\}$ then $A^c$ is not in $\mathcal{G}_{\sigma c}$ as only $A \cup B$, $(A\cap B)^c$ are in $\mathcal{G}_{\sigma c}$. Am I correct? or am i missing something?