Existing words already have meaning. The relationship between the word (form) and the meaning is conventional, and in learning a language (as a child or second language learner), you learn that relation. This is a very complicated topic, so will just say that "you learn" without going into all of the details of how you learn, but children learn inductively by associating forms and contexts, which allows them to internalize a definition. This gives an individual meaning.
Because of the predominantly social view that most people have of language, we also have to relate individual meaning with group meaning. For natural objects, there isn't much difference between the two. However, there is a kind of skillet known in parts of eastern US as a "spider", where some people know that additional meaning of the word but most people do not. We then talk about dialectal complications. Less obscurely, "boot" has a different meaning in US English compared to UK English. It should not be hard to understand the complexity of standardized dialect differences, as long as you don't insist that the form-meaning relation must be rigid within a language (in the broadest sense).
It is beyond the ability of linguistic science to give a physical description of "meaning" as an aspect of the mind (whereas physicists can say what an "atom" is, and not just what it does). It is obvious that the meaning of a word is not the sum of experiences that a person has had with a word (word meaning would then be very unstable even within an individual), but it is probable that meaning is a mental abstraction based on such experiences. Operationally, we can say that for two individuals the meaning of word A is the same iff the word picks out the same referents. (However, borderline cases cause a problem for that even within an individual).
Looking word meaning up in the dictionary is the other main method of acquiring meaning. I had to look up "squamulose", because I couldn't figure it out from context (I also didn't know the word "squamata"). In a good dictionary such as the Oxford English Dictionary you will get a more complete descriptive account of the meaning of words, which is based on observation. Not all dictionaries are good.
Individuals may deliberately create new form-meaning pairs, perhaps by taking an existing semantic description (maybe with a nuance) and mapping it to a new form, which leads to the peculiar situation that it's good to be bad, or sick. One might create a new object and then assign a name to it (a daily event in the pharmaceutical industry).
Whether or no such a new form-meaning association will generally be accepted is pretty much unpredictable. If you can get a pop culture icon to use your new form-meaning pairing, you have a greater probability of gaining somewhat widespread social acceptance. Figuring out the actual social dynamics of how this happens is very difficult, since we generally don't have very complete records.
The attitude of the individual who initially creates a form-meaning association is mostly irrelevant, what matters is whether that association is widely-enough disseminated. Explicit rules may exist in academic publications, especially in the formalistic sciences, but usually the form is an existing word (or modification of an existing word). Our experience with linguistic terminology is that the form is most solid and the meaning is highly flexible, for example "phoneme", "markedness", "constraint". In a formal academic paper, an author may indeed lay down a rigid definition, and it may have a live span of three or four years, then the edges get smoothed down and instead we have a collection of meanings.