First suppose R=kG is von Neumann regular. As a quotient ring of R, k is certainly von Neumann regular. Consider any finite subgroup E⊆G, say of order m. It suffices to show that any prime p∣m is a unit in k. To see this, we fix an element σ∈E of order p (which exists by Cauchy's Theorem). Taking an element α∈R such that 1−σ=(1−σ)α(1−σ), we can argue that p∈U(k). Finally, let H be a subgroup of G generated by a finite set of elements, say h1,…,hn. It is easy to see that
I:=h∋∈H∑R⋅(h−1)=i=1∑nR⋅(hi−1)
(using facts such as hi−1−1=hi−1(1−hj), and hihj−1=hi(hj−1)+hi−1). Since R is von Neumann regular, I=Re for some idempotent e. Now I is contained in the augmentation ideal of R, so e=1 (since k=0). Thus, f:=1−e=0. But for any h∈H, (h−1)f∈I⋅f=R⋅ef=0⇒f=hf. Since f involves only finitely many elements of G, this implies that ∣H∣<∞.
For the converse, let us assume the given conditions on k and G. Consider any element α=a1h1+⋯+anhn∈R (ai∈k,hi∈G), for which we want to show α∈αRα. Let H be the (finite) subgroup of G generated by h1,…,hn. Since α∈kH, we are done if we can show that S:=kH is von Neumann regular. Consider any principal left ideal, say S⋅β, where β∈S. Viewing S⋅β⊆S as k-modules, we have S∼k∣H∣ and S⋅β=∑h∈Hk⋅(hβ). Since k is von Neumann regular, S⋅β is a direct summand of S as k-modules. Then S⋅β is a direct summand of S as S-modules. This checks that S is von Neumann regular, as desired.