A feasible "proof of brain" can be implemented until the cost in time and engagement would invalidate most of the attempts of manipulation. Security always is enforced by rising the effort and lowering the rewards.