yes, that should be the case, except what if an agent is running on an AI server, & carrying out instructions from the inference server, directly from the AI server? Also, what about conspiracy laws, where the inference server is actively involved in planning the crime?