We can formulate your issue as:

Minimize $\|Ax\|$ under the constraint $\|x\|=1$

than can be re-phrased in the following equivalent way (because it is the Euclidean norm)

Minimize $\|Ax\|^2=(Ax)^T(Ax)=(x^TA^T)(Ax)=x^T(A^TA)x$ under the constraint $\|x\|=1$

As a consequence, it is the eigenvectors of $B:=A^TA$ that must be considered.

(have you heard about SVD and singular vectors and singular values ?)