Let $P$ be our candidate for the "closest point", $A$, $B$, $C$ the vertices of the triangle and $M$ the circumcenter of the triangle.
Let $d_A = \|A-P\|,$ $d_B = \|B-P\|,$ $d_C = \|C-P\|,$ $d_M = \|M-P\|.$
Let $R = \|A-M\| = \|B-M\| = \|C-M\|.$
With some tedious calculations, we find
$$
\frac{1}{3} \left(d_A^2+d_B^2+d_C^2\right) = d_M^2 + R^2
$$
(holds in equilateral triangles, only). This obviously means that the largest of the three values $d_A^2,$ $d_B^2,$ $d_C^2$ is greater than or equal to $d_M^2+R^2.$ If they all were smaller than $d_M^2+R^2,$ the equation could not hold. Therefore, $\max(d_A,d_B,d_C)\geq R,$ independently of the location of $P$. But we already know one configuration with $d_A=d_B=d_C=R,$ specifically the one in which $P=M.$ As we have shown that we cannot do better than this, this is the optimal solution.
We even know that this solution is unique. The equation above shows us that $d_A=d_B=d_C=R$ cannot hold if $P\neq M,$ because $P\neq M$ implies $d_M^2\neq 0,$ which in turn implies that one of $d_A^2,$ $d_B^2,$ $d_C^2$ must be greater than (not equal to) $R^2.$