You are safe to publish the resulting vector graphics.
Please confirm that you generate the graphics as follows:
From what you said, I assume you are using the statistics library and its boxplot plot handler as in the following example:
\documentclass{standalone}
\usepackage{pgfplots}
\usepgfplotslibrary{statistics}
\begin{document}
\begin{tikzpicture}
\begin{axis}[y=1cm,ytick=\empty]
\addplot+[boxplot]
table[row sep=\\,y index=0] {
data\\
1\\ 2\\ 1\\ 5\\ 4\\ 10\\
7\\ 10\\ 9\\ 8\\ 9\\ 9\\
};
\end{axis}
\end{tikzpicture}
\end{document}

In this case, pgfplots computes a boxplot prepared out of the input data. The values of this boxplot prepared are (a) visible and (b) can be reverse engineered. But the original table values are unavailable from the resulting vector graphics (except for any outliers, of course).
Unrelated side-remark:
Note that this answer holds only for boxplot. If you had a scatter plot or a line plot, the entire coordinate data could be reverse engineered. In fact, the clickable library of pgfplots allows not only to reverse engineer any coordinate, but also contains a copy of the input coordinates in the resulting .pdf.
standaloneand include it as an image? – Ethan Bolker Sep 13 '13 at 12:22