This is a pretty silly claim.
The first reason is pretty obvious: If it’s in digital text, then since the underlying character values are all the same as normal text and the font is irrelevant, well, the font is then irrelevant.
But for images or printed copies on paper, the only thing needed is for the OCR software to recognize the font and then match the glyphs (characters) to the proper letters, and again the font is irrelevant. Perhaps it might take the NSA a while to clue into it, but they’d eventually “get it”. Given that they store so much data, how long it takes for them is almost irrelevant.
It’s all pretty simple. Which makes me wonder whether the Reason.com article is purposeful disinformation or whether it is simply not well thought out by someone who isn’t very techno-savvy. It’s probably more likely that it’s just not well thought out, but given the simplicity of the issue, that’s somewhat hard to believe, leaving the question of whether it is purposeful disinformation open.