Beyond Bias and Compliance: Towards Individual Agency and Plurality of Ethics in AI