DataCite exporter double-escapes HTML in description metadata

Problem description
HTML markup in the “Description” field is xml-double-escaped when exported to Datacite.

As a result, HTML content is displayed literal in Datacite Commons and, consequently, in OpenAire as well when OpenAIRE harvests metadata from DataCite.

See https://commons.datacite.org/doi.org/10.7910/dvn/glqwah
In this example descritpion, the linebreaks are rendered as:
... women. <br /><br /> The sample ...

However, a line break is expected.

Here is the cURL request to view the metadata:
curl "https://dataverse.harvard.edu/api/datasets/export?exporter=Datacite&persistentId=doi:10.7910/DVN/GLQWAH"

And here's the response:
... women. <br /><br /> ...

But it should be:
... women. <br /><br /> ...

On the other hand, here's a "working" example from Zenodo that is displayed in DataCite Commons using interpreted HTML, where the text is correctly displayed in bold:
https://commons.datacite.org/doi.org/10.5281/zenodo.6861794

The description in the correspondending downloaded example DataCite XML is not double-escaped:
<description descriptionType="Abstract">&lt;p&gt;&lt;strong&gt;Abstract: &lt;/strong&gt;Feudalism

That's why I suspect that escaping only once would solve the problem. Unfortunately, I can't really test it, so I'm not sure.

Affected versions
at least 6.10 and 6.11 (and probably older versions as well)

Previous Discussion
https://groups.google.com/g/dataverse-community/c/dXqHmgpdJ9k

Any open or closed issues related?

• https://github.com/IQSS/dataverse/issues/12625

Start by reproducing the Datacite exporter through the mentioned /api/datasets/export?exporter=Datacite endpoint using the provided persistent ID, then trace how the Description metadata is escaped. Compare the current response with the expected single-escaped form and verify that HTML line breaks remain interpretable without affecting other metadata.