In recent weeks, a curious phenomenon has circulated on social media: Google Photos users reported that images of children appeared transformed into oil portraits or hand-drawn sketches. The most widespread explanation pointed to an artificial intelligence filter, but the debate soon shifted to a deeper issue: is Google using our photographs to train models like Gemini? For many, the answer is uncomfortable. What appears to be a simple visual effect reveals a chain of technical and commercial decisions that directly affect privacy and data control.
To understand the real scope of this situation, it is best to step back from the anecdote and look at the entire ecosystem. Every photo we upload to a cloud service — whether Google, Apple, or any other — goes through an automated analysis process: facial recognition, object tagging, scene classification. These operations require artificial intelligence models that, in turn, need enormous volumes of data to train. The line between functional processing (like searching for "dog" in your gallery) and training proprietary models is much thinner than most users believe. And that is the real problem: we do not know for certain what use platforms make of our images once they have been analyzed.
The scenario that emerges is not a conspiracy, but a common practice in the technology industry. Companies optimize their cloud services AWS and Azure to manage petabytes of visual data, and AI training is one of the main economic drivers. However, when that data comes from users who have not given explicit and informed consent (beyond a checkbox in the terms of service), ethics wobbles. The reality is that services like Gemini, or any other generative model, may be feeding on a gigantic photographic catalog without the owners of those images knowing. And that, from a corporate standpoint, is a major reputational and legal risk.
Faced with this landscape, companies that handle sensitive data need to rethink their information management strategy. Many organizations still use consumer solutions to store product photos, internal documentation, or even customer data. This practice, although convenient, exposes the company to having its visual assets end up as part of third-party training sets. That is why more and more businesses are opting to develop custom applications and custom software that allow them to retain full control over their data, from storage to processing. At Q2BSTUDIO, we understand that cybersecurity and data sovereignty are fundamental pillars, and that is why we offer solutions that integrate AWS and Azure cloud services with customized privacy policies, preventing critical images or documents from leaving the organization's controlled perimeter.
Furthermore, when implementing business intelligence services like Power BI, it is crucial to ensure that source data — including images — is processed under anonymization and consent criteria. AI for business does not have to be based on stolen or questionable data; it can be trained with its own, auditable, and internally managed datasets. Even AI agents that automate workflows can be designed to work exclusively with corporate information, without relying on external APIs that analyze photos on third-party servers. This not only protects privacy but also builds trust among customers and partners.
Returning to the initial debate, the accidental transformation of children's photos into paintings is just a symptom of an opaque system. The conversation should go beyond the filter and focus on how companies — large and small — manage the treasure trove of visual data they accumulate. The solution is not to stop using the cloud, but to do so with intelligence, transparency, and the right technical tools. At Q2BSTUDIO, we help design and implement those tools, from artificial intelligence platforms tailored to each business to cloud storage systems with granular access controls. Because the best way to prevent your photos from feeding third-party models is to build your own digital ecosystem.
