Generally speaking JMP reports statistical inference statistics such as Prob > F ratios, Prob >|t|, etc. with p values which you can use to compare to any significance level threshold you want use. So there really isn't a significance level default value for various inferential decisions inherent in JMP's modeling platforms.
Peter is correct in his statement, but I am looking at your question in a different way. You can use stepwise regression and set the level of significance you are willing to accept for being right/wrong that a predictor is important or not to your model. That would be set under stopping rule for P-value Threshold. A rule of thumb is to set that value at 0.05 for leaving. This translates into all of your important factors needing a P >|t| of 0.05 or less if they are to be included in the final model. You can change that significance value to any value you choose, but I would recommend keeping it as low as possible.