How so? You'd have to figure out the N in a population of hundreds of thousands (the state's prison population) without any other variables to go on.

It breaks down the N by location, which... what defines a location here? Is it a state? Is it a single prison? For a small N, narrow location is definitely a de-anonymization factor.

There's no mention of any narrow locations. The numbers are at the state level per the request in the article.

How would knowing "1 Transwoman is currently housed in a female prison on the territory of Washington state" in any way risk causing harm to that transwoman, or help you identify her in any way?

It doesn’t. I think GP is confusing small cell counts in micro data (a real problem) with small numbers in aggregate data (sometimes a problem).

In this case, just a count and prison is not a privacy issue as someone would have to already know the individual is trans to identify them. And that’s the only information contained in the data release.