Evaluering af AI-modeller under nye standarder