This one is very borderline and I certainly won't be offended/surprised if it gets closed. [EDIT: I clearly misjudged this---the question has a good few upvotes now.]

The London Maths Society (LMS) are thinking of (indeed, actively pursuing the idea of) reprinting "Algebraic number theory" Ed. Cassels and Froehlich. Hence the LMS had to go about contacting the authors of the original articles. When they contacted Serre he replied "sure reprint my articles, but please include the erratum that I indicated in my completed works." Sure enough, he had found a slip in one of his articles (in the statement of the "ugly lemma"---Serre went on to say that this had taught him not to abuse lemmas, as they might bite back!) and had taken the trouble to fix it when his completed works were published.

I looked over the thread from last October about the errata database but the database doesn't seem to contain this book. On the other hand it is surely a very widely-read book. I think I just found a typo in the definition of a co-induced module on p98: I think the $G$-action on $Hom_{G'}(\mathbf{Z}[G],A')$ ($G'$ a subgroup of $G$ and $A'$ a $G'$-module) should be that $g\in G$ sends $\phi$ to the function sending $g'$ to $\phi(g'g)$, not what the authors say (what they say isn't even an action as far as I can see, unless I made a slip). The notation is also terrible: $g'$ is in $G$, it seems to me.

Does anyone else have any scrawled marginal notes in their copies of Cassels-Froehlich about typos or other things that the LMS can fix? They are planning on having an erratum page at the beginning of the book when they reprint it.

EDIT: someone from the LMS got in touch with me to say that the only errata they would dare publish would be errata that had been confirmed by the authors, or someone "of a similar standing", so in fact it might be the case that not everything mentioned here gets put in the LMS erratum.

If this thread does get closed, feel free to email comments to buzzard at imperial.ac.uk.

Edit: Discussion on this thread is happening here at the meta.

Edit: the LMS have set a deadline of 1st Feb. After this point I (KB) will put all the errata we have caught into one file and the LMS will send it to the authors, asking for their approval.

Edit: Anton and Ilya have suggested that really this would be better if it had one big answer rather than lots of smaller ones. But let me persist with the "lots of smaller ones" for the time being, because I am still getting emails with non-trivial lists in from different sources and, although I want to put everything together into one pdf file, I don't really want to do it until I am pretty sure no more is coming in. On the other hand these partial lists have definitely been of help to some people, e.g. I've had emails saying "I have a big list of corrections; here are the ones that haven't already been mentioned."