[guided]We assume that the single equation holds at the fixed integer tuple $a$, namely
\begin{align*}
G(a_1,\ldots,a_n)=0.
\end{align*}
By the definition of $G$ and by the notation $b_j:=F_j(a_1,\ldots,a_n)$, this is the same as
\begin{align*}
\sum_{j=1}^{m}b_j^2=0.
\end{align*}
The key point is that the summands are not arbitrary integers; they are squares of integers. For each $j\in\{1,\ldots,m\}$, we have $b_j\in\mathbb{Z}$, so $b_j^2\in\mathbb{Z}$ and $b_j^2\ge 0$. Thus the displayed sum is a finite sum of nonnegative integers.
We now prove that no nonzero $b_j$ can occur. Suppose that there is some index $k\in\{1,\ldots,m\}$ such that $b_k\ne 0$. Since $b_k$ is an integer, the condition $b_k\ne 0$ implies $|b_k|\ge 1$. Squaring gives
\begin{align*}
b_k^2\ge 1.
\end{align*}
Every other term $b_j^2$ is nonnegative, so the full sum satisfies
\begin{align*}
\sum_{j=1}^{m}b_j^2\ge b_k^2\ge 1.
\end{align*}
This contradicts the assumed equality $\sum_{j=1}^{m}b_j^2=0$. Therefore no such index $k$ exists, and hence $b_j=0$ for every $j\in\{1,\ldots,m\}$.
Substituting back the definition $b_j:=F_j(a_1,\ldots,a_n)$, we obtain
\begin{align*}
F_j(a_1,\ldots,a_n)=0 \text{ for every } j\in\{1,\ldots,m\}.
\end{align*}
Thus any integral solution of the single equation $G=0$ is an integral solution of the original system.[/guided]