2-sets exist in \(\mathbb{R}\): for example, a set which intersects each line exactly at two points. It can be constructed using transfinite recursion over first uncountable ordinal, using the fact that each line in \(\mathbb{R}\) can be uniquely represented as a triple of reals, and finite cartesian power of infinite set has the same cardinality. Hence, a set of all lines on the plane is uncountable, and, using Zermelo's theorem, can be well ordered.
The set will be constructed inductively: suppose that for all ordinals \(M_{\alpha}, \alpha < \beta\), the following invariants hold: no three points lie on the same line and for each \( l_{\alpha} \) there are exactly two points that intersect it. Consider a union of all \(M_{\alpha}\): the invariants still hold (as otherwise, the problem would arise at some earlier step), and, if \(l_{\beta}\) intersects this union at two points, we are done and \(M_{\beta} = T\), otherwise we should choose two points in a manner that doesn't violate the invariant for \(T\).
How can we proceed? First of all, there are \(\le 2\beta\) points already chosen, and, as ordinals, \(2\beta = \beta\), and a set of prohibited directions has an order type \(\beta^2\), and it has strictly smaller cardinality. Hence, as \(M_{\beta}\) has countable cardinality, there always remains enough points to choose from (simply choose two points lying on \(l_{\beta}\), such that the pair doesn't belong to the set of prohibited ones).
The very same process allows one to construct a set of points of a sphere which intersect each big circle exactly at 2 points.