Yes, there are data privacy considerations, but the main ones called out in the docs are:
Spark uses a generative AI service (OpenAI). [Spark docs]
By default, Mixpanel does not use your customer data to train the AI model; you can optionally enable training on your data in settings. [Spark docs]
The knowledge sources do not explicitly mention different default behavior for EU vs US accounts (such as Spark being auto-disabled for EU orgs), so I can’t confirm that part from the docs alone.