Venice uses different Llama3 versions on their servers so privacy has trust involved. If you want sure more privacy you must run the models local.