The abstract of an article published in 2009 (Detection of Steganography Inserted by OutGuess and Steghide by Means of Neural Networks) reads:

Results in this paper show that used models had almost 100 % success in steganography detection of messages inserted by OutGuess and Steghide.

I've not read the paper, but steghide seems not maintained upstream since nearly ten years, and I've not found any kind of follow-up to this article. Personally, I'd need quite more certainty about the level of safety offered by steghide before we seriously consider shipping it to Tails users.

In any case, if you insist to use steghide now and take the risk, it's in Debian and Tails is Debian based, so you can (unsupported) install it yourself.

Okay, thank you.

Results in this paper show that used models had almost 100 % success in steganography detection of messages inserted by OutGuess and Steghide.

That is terrible. Is there another way, to send messages in a hidden file, witout detection?

According to what a researcher in steganography recently told me in private, there seems to be no existing tool up-to-date with the attack / defense research.