P

The Dilema

There seems to be a dilemma facing creators about the rise of generative Ai. It seems to encompass a few things at once. One, a sense of economic threat to artists bottle line. The idea that machines will replace human labour with Ai, that artistic works will become commodified, debased and devalued. Two, a lack of consent. Ai companies who used artistic work to train models without compensation - theft of artitist work that contiunues as millions use generative AI. Third, a fundamental power struggle over artistic integrity. The asymmetry between up-and-coming or establish aritists and AI corporations - the creators vs the sellers.

But first, let's narrow down some terms. Specifically, generative AI. First, let's look at the broader field of Artificial Intelligence, which exists in a larger hierarchy of AI methodologies. Artificial Intelligence (AI) is a member of the broader field of computer science, focused on enabling machines to mimic human behaviors, such as visual perception or decision-making. Machine Learning (ML) is a subfield of AI that focuses on how computers can lean patterns without being explicitly programmed for every scenario. Its focuses on training through numerous examples. Deep Learning (DL) uses neural networks to learn complex patterns. Lastly, Generative Ai (Gen Ai) is a subset of Deep Learning and uses many difference techniques, such as Variational Autoencoders (VAEs), Generative Adversarial Networks (GANs), and Diffusion Models to create what we consider Generative AI content.

Economic Threat

For many creators, the economic threat of Gen AI, focuses on the loss of revenue from future commissions, devaluing existing economic investments, and unresolved compensation for use of copyrighted works. Creators report witnessing a loss of clients and commissions to AI platforms that generate good enough quality for cheaper. The additional competition from AI companies trained on models that reiterates their work for free, have many harmful effeects. The legal cost of enforcing intellectual property (IP) protection for artists of all sizes is large and drawn out. There have been those who have taken on the burden of filing copyright infringment claims, as in Andersen v. Stability AI Ltd., For others, it is too costly to fight if a copyright infringment and monitor for infringements.

We may have to define the term creators, for prosperity.

Lack of Consent \

VODE aims to help.

Similar to copyright

Artistic Integrity

Ex: Garage band

Ex: Prosumer and smart phone photography

The conflict

The Question

Adversarial Ai

The conflict of AI scraping is an inportant one. Similar to how network firewalls become a requirements of packet switched networks, so will the need to adopt sophisticated defensive tools to protect and enforce intellectual property rights.

What is it and does it work?

Adversarial AI techniques can be applied to images posted online, including photography, but the effectiveness depends heavily on the goal and the countermeasures in play. Let’s unpack it in practical terms.

Is the focus too narrow on "capturing" profit from those who profit?

What if we are protecting "medical records" or something equally sensitive, from AI scrapping.

There must be allowed passive-reactive defensive from copyright holders against the inadvertent negligence of AI privateers.

If we’re talking about adversarial attacks—subtle pixel-level perturbations designed to fool AI models—they can be embedded in photos. These modifications are usually imperceptible to the human eye but can cause machine vision systems (like facial recognition, object detection, or AI image classifiers) to misidentify or fail to detect what’s in the image. This approach is called adversarial perturbation or adversarial noise injection.

However, for photos posted online, three major caveats limit their real-world use:

That said, there’s a related and growing field called “adversarial cloaking” or “data poisoning for privacy”. Tools like Glaze or Nightshade (from the University of Chicago) intentionally add protective noise to artworks and photos, aiming to mislead AI training systems that might scrape them. Instead of tricking classifiers, they corrupt the training signal for generative models.

So yes—adversarial AI can work on online photography, but only if it’s carefully engineered for a known target or designed for anti-scraping use cases. It’s less “upload this and fool all AI” and more “strategically distort your content to protect it from specific machine vision tasks.”

Would you like me to walk through how photographers are currently using cloaking tools to protect their work from AI datasets?