It is wrong to treat people in any way to which they would not have sufficient reasons to consent in the act-affecting sense. [To consent in the 'act-affecting' sense is not to perform a speech act, for which there might be confounding instrumental reasons, but just to have the attitude whereby if you had power over the person's choice, you would allow them to act in this way.]
He then considers the objection that this principle is superfluous (even if true):
[W]hat is morally important, these writers claim, is not the fact that these people could not rationally consent to these [wrong] acts, but the various facts that give these people decisive reasons to refuse consent.
As an example, Parfit imagines an Act Utilitarian who claims:
(I) Everyone could rationally consent to all and only the acts that would, on the whole, benefit people most.
However, rather than making the Consent Principle superfluous, Parfit suggests that the truth of (I) would instead show that the Consent Principle provides independent support for Act Utilitarianism:
(I)'s truth would give us a further reason to believe that these acts were morally required, and a further reason to act in these ways.
This seems to me mistaken, or at least arguably so. Parfit holds that "if there is only one possible act to which everyone could rationally consent, this fact would give us a strong reason to act in this way" (emphasis added). But this is just what the imagined objector is denying. It would be most natural for Act Utilitarians, especially, to think that these facts about hypothetical rational consent do not give us additional reasons. Instead, what they do is signal (or even guarantee) that there are other reasons for us to choose this act over the alternatives. (In particular, it signals that this is the act that would, on the whole, benefit people most. And it is this latter fact, and not its signal, that provides us with reason to so act.)
Parfit's view risks double counting, especially if we think that (I) is true because Act Utilitarianism is true. Imagine someone arguing as follows:
(1) Everyone could rationally consent to all and only acts that are, on prior grounds, morally permissible.
(2) AU is true: permissible acts are just those that would benefit people most.
(3) Therefore, (I) is true.
# (4) Therefore, we have additional reason to act in these ways.
Such reasoning seems dubious. In order to defend the non-superfluity of the Consent Principle, Parfit really needs to argue that my premise (1) is false: it must be that, though the two coincide, the facts about rational hypothetical consent are at least partly 'upstream', rather than entirely 'downstream', of the moral facts. That's a defensible view (indeed, I take it this is the basic idea behind Scanlon's Contractualism). But I don't see that Parfit has said anything here to defend it -- let alone to suggest that Act Consequentialists are committed to it.