RAIS_WG aims to define specifications to guide developers in building AI-based systems, including image recognition systems. These specifications are set out to test for accuracy, ensure resistance against malicious attacks by third parties, and ensure that the system cannot be easily fooled by adversarial attacks that expose its weaknesses. Facial characteristics within images can be subject to racial and gender bias from image recognition systems, which can compromise an individual’s ability to use various software and applications. Moreover, the normalisation of facial recognition use may further facilitate the biometric mass surveillance of people and communities.