Based on case law, there seem to be two contradictory tests for valid prior art when considering novelty of an invention. I would paraphrase them as:
(i) If information was made accessible and copyable to (even a limited) public audience with sufficient expertise and without explicit/implicit expectations to not copy/distribute the information (i.e. not trying to 'keep secret'), it is considered valid prior art. (See Rosaire v Bariod 1955, Klopfenstein 2004.)
(ii) On the other hand, if a printed publication was not indexed by subject matter and not reasonably possible for an interested researcher to find, it is not considered prior art. (I don't have a specific case for this, but the precedent of a manuscript filed with and publicly accessible via the federal govt, though not indexed by subject matter was highlighted as an example of non-printed publication in this video.)
The contradiction comes about because if information is disclosed to an arm's length party as in (i), that party may not distribute/index the material, and thus it would not be reasonably possible for an interested researcher to find unless he interviewed every person on the planet. So if an invention was disclosed 'publicly', but was not publicly disseminated or indexed such that it was searchable, would an uninformed third party who independently came up with the invention still have the ability to acquire a patent if it filed first?
If the non-disseminated disclosure is considered 'public', then this video states that it would disallow others from patenting. The video also notes that a more apt description for post-AIA is 'first-to-disclose' for this reason.