In contemporary digital society, facial images possess profound operative significance as artefacts conveying information about individuals, their emotions, and their identities. They operate as proxies of identification processes intricately interwoven with the material dimension of society, where tangible artefacts such as photographs, datasets, and technologies for automated recognition play a pivotal role. As proxies, images operate as identity markers and catalysts for recognition and identity construction. They are not static representations but dynamic agents that shape our perception, guiding us in assimilating characteristics from the world around us. Within this framework, my presentation will explore CAMOUFLAGE, an image anonymisation technique utilising Latent Diffusion Models (LDMs) developed by Lia Morra's team of computer scientists at the Polytechnic of Turin, Italy, as part of the outcomes of the ERC Project FACETS led by Massimo Leone. This method distinguishes itself from previous approaches, such as Generative Adversarial Networks, by anonymising entire images, including faces, bodies, and backgrounds, while retaining critical analytical features. This approach generates valuable synthetic data for social research, offering anonymised yet insightful datasets. Considering this technology, I will consider gender identity as a focal point for thinking about images as proxies of identity. The following research questions propel my paper: How does gender function within an anonymisation process? If, as feminism and cultural studies have taught us, gender is a performative aspect of our existence rather than a quantifiable datum, how can it become operational?