How can AI detect harmful, abusive, or inappropriate content on platforms like OwnMates while preserving users’ freedom of expression? What AI techniques can help distinguish genuinely harmful content from opinions, criticism, humor, or cultural differences?