Article 27 requires deployers of high-risk AI to assess fundamental rights impacts before deployment.