Here are some observations that are slightly too long for a comment.  First, note that the claim is false if we replace $5$ by $4$.  

**Claim.** There exists a graph $G$ with $\chi(G)=4$ such that $G$ does not contain a $3$-connected subgraph.  

*Proof.* The [Moser Spindle][1] is a well-known graph with chromatic number $4$ (since it is obtained via the [Hajo's construction][2] applied to two copies of $K_4$).  Moreover, it is easy to see that no subgraph of the Moser Spindle is $3$-connected.  

**Claim.** Every graph $G$ with $\chi(G)=5$ contains a subgraph $A$ and $x,y \in V(A)$ such that $A+xy$ is $3$-connected.  

*Proof.* Let $G'$ be a subgraph of $G$ such that $\chi(G')=5$ and $G'$ is minimal (under the subgraph relation) with this property.  If $G'$ is $3$-connected, we are done.  Otherwise, let $(A,B)$ be a $(\leq 2)$-separation of $G'$ where $A$ corresponds to a leaf vertex in the [SPQR tree][3] of $G'$.  By the minimality of $G'$, $\chi(A), \chi(B) \leq 4$.  Thus, $(A,B)$ is a $2$-separation and the vertices $x,y \in V(A) \cap V(B)$ must be non-adjacent in $G'$ (otherwise $G'$ is $4$-colourable).  Moreover, we must have $\chi(A)=\chi(B)=4$ (otherwise $G'$ is $4$-colourable).  By the definition of the SPQR tree, $A+xy$ is either a cycle or is $3$-connected.  However, $A+xy$ cannot be a cycle since $\chi(A)=4$.  Thus, $A+xy$ is a $3$-connected, as required.  


  [1]: https://en.wikipedia.org/wiki/Moser_spindle
  [2]: https://en.wikipedia.org/wiki/Haj%C3%B3s_construction
  [3]: https://en.wikipedia.org/wiki/SPQR_tree