That alone would not be sufficient though, you could have learned about it by chance - it would be more useful as a meta-encoding, like a watermark, with multip

That alone would not be sufficient though, you could have learned about it by chance - it would be more useful as a meta-encoding, like a watermark, with multiple 'trap concepts' that are connected enough with everything the model knows as well as 'white spots' at certain points, e.g. about a fictional book or movie. Edit: so the existence/absence of ctrap concepts encodes the 'version'.