Garrett Cullity, in The Moral Demands of Affluence, argues against extremely demanding conceptions of beneficence by appeal to the following principle: "Someone else's interests in getting what it is wrong for her to have cannot be a good reason for requiring me to help her." (For example, we obviously shouldn't help a gangster unjam his gun and shoot his victims, in the absence of any other reason to do so. His interests in wrongdoing don't suffice.) Let's call this the 'No Helping Wrongdoers' principle, or NHW for short. In this post, I'll briefly explain how Cullity uses NHW to argue against extreme demands of beneficence, and then I'll show that NHW is actually false. The intuitive cases appealed to in its support really only support a weaker principle that is insufficient for Cullity's purposes.
Cullity's argument: We can describe a scenario in which you would be required by beneficence to advance Bob's (non-altruistic) personal interests, say in developing his musical talents. By NHW, we only have such reasons if Bob's interests here are morally permissible. So it must be permissible for Bob to develop his musical talents (even though he could be saving lives instead). To generalize: morality isn't generally extremely demanding, or it would - by NHW - absurdly imply that many ordinary personal interests fail to provide us with reasons of beneficence. (It would, in a sense, be insufficiently demanding in ordinary contexts.)
Why NHW is false: Gangster examples support the narrow claim that we shouldn't help people in ways that make things worse overall. But NHW goes further than that, as cases like Bob's show. The problem is that some intrinsic good X might be "wrong" to pursue for the purely comparative reason of opportunity costs: though good in itself, there's something else you could be doing that would be even better. But suppose that you just can't bring yourself to make the required sacrifice to achieve this impartially best outcome Y. Further suppose that you might even fail to achieve X, in which case the result would be the dreary outcome Z, containing neither personal nor impartial goods. Now let's bring a third party into the picture. Suppose I'm in a position to easily help you achieve X, rather than Z. Should I do so? Of course! Better X than nothing. But according to NHW, I should not help you here, since (we've stipulated) your interest in X is one that it's wrong for you to achieve, given that you ought instead to have pursued Y. We thus see why NHW is false. It fails to recognize that we sometimes ought to help someone to achieve a lesser good, if the realistic alternative to our aid is that there will be no good done at all.