The guideline also mandates that AI systems be “understandable by design”, allowing regulators and users to clearly assess their functioning, and calls for grievance redressal mechanisms to prevent bias, discrimination, and misuse, including deepfakes.