This example does not involve Discourse, but the policy and governance issues would certainly overlap.
Both the English and German versions of Wikipedia now have what I would describe as extreme AI-purity policies. As an editor with 16 years experience, I have been wrongly accused of adding AI‑generated content to an article that I predominantly wrote. And the process of seeking a remedy has been bruising. That includes filing formal complaints to the Wikipedia EN ArbCom committee and the Wikimedia Foundation Trust and Safety Team (WMF TST) — dismissed summarily or rejected without reason, respectively, thus far.
I don’t want to recount my experiences here, partly because they are ongoing. However, I would encourage readers here to review this extraordinary exchange regarding the Wikipedia EN article on Fern Cave, a site of cultural significance to the Modoc people of California:
The author in question has to deny using AI four times. Their indigenous knowledge is also discounted in the process. The enforcing editor concludes with a remark that resembles concern but that can equally be read as a veiled threat to sanction if they fail to yield.
Psychological tactics like these are regularly deployed by enforcing editors and I have evidence of other examples.
If you look at the source markup in that same exchange, you will find a font-size: 0pt honeypot trap set to try and detect AI usage by the accused author when formulating replies. This trap was not sprung.
There are credible reports of several editors abandoning Wikipedia DE after being “unfairly attacked” over their alleged use of AI tooling.
To answer the original question posed by this topic, I would argue that Wikipedia EN and DE are failing badly in their governance in multiple ways in their attempts to deal with AI content. I documented three case studies in my recent 17 page report to the WMF TST team. I may well make that PDF public in due course?