General Discussion

I would be careful with using regex for content moderation (see the Scunthorpe problem).

There are open source machine learning models that could be used to detect toxic content. See

• https://huggingface.co/unitary/toxic-bert
• https://github.com/unitaryai/detoxify

I dont know how much compute power you would need to process the content of a lemmy instance using this. But it would be more reliable than regexes.