Understanding AI training and protecting your artwork

In recent years, artificial intelligence and machine learning have made significant strides in various fields, including the world of art. AI models now capable of creating, analyzing, and even mimicking artistic styles by learning from vast collections of existing artworks. But with this controversial technology comes a big question: how can real artists protect their original creations from being copied or misused by AI?

Let’s dive into how AI training works, how software companies are using it to develop generative AI tools, and—more importantly—what you can do to keep your art safe in this digital age.

What is AI training?

Imagine you’re learning a new skill or starting a new career. You’d probably look into what others in the field are doing, you’d look into their workflows, and try out tried and tested techniques. AI training works similarly. It's a computer learning to recognize patterns and make decisions based on what’s worked in the past.

In more technical terms, AI training involves feeding a computer algorithm a large amount of data, which it uses to learn how to perform a specific task. For instance, if someone is training an AI to recognize different styles of art, they might show it thousands of paintings from various artists and periods. Through analyzing these artworks the AI grasps patterns, colors, brush techniques and other characteristics that distinguish each style.

This process involves a few key steps:

1. Data collection: Gather a large number of examples for the AI to learn from.
2. Training: Run this data through a machine learning model, which adjusts itself to better recognize patterns and make accurate predictions.
3. Validation: Test the trained model with new, unseen data to ensure it has learned correctly and can apply its knowledge to different situations.
4. Iteration: Repeat the process, tweaking the model and the dataset to continually improve its performance.

The goal is for the AI to become so good at identifying these patterns that it can, for example, look at a new piece of art and tell you which artist's style it mimics, or even create a new piece of art in a particular style.

How software companies integrate AI into their products

AI uses different types of models to work with art. Some create new artworks, others analyze and classify images, and some improve the quality of images. For example, models like GANs (generative adversarial network) and VAEs (variational autoencoder) make new art by learning from existing pieces, while CNNs (convolutional neural networks) help recognize styles and subjects. Stable Diffusion improves image clarity, and Transformers generate detailed art.

These AI models help design software companies like Adobe, Canva, and Linearity deliver powerful features that make designing both easier and more efficient. While these advancements in AI open up new opportunities for designers, they also pose significant challenges, particularly regarding the protection of artists' work. While most companies, including Linearity, use AI models that use data sourced from publicly available images, licensed content, or user-generated inputs, others take a less ethical approach.

Meta, for example, recently faced backlash for collecting user data without explicit permission. Many artists, feeling frustrated with their data being used for AI training, are leaving Instagram in large numbers. Some are now switching to Cara, a crowdfunded, anti-AI social media and portfolio platform.

This shift underlines crucial issues regarding copyright and the imperative for artists to safeguard their work from AI and unauthorized use. But don’t worry—In the following section, we’ll unveil our tips on how you can effectively protect your valuable work.

How to protect your work from AI

Thankfully, there are several effective ways to protect your artwork from AI misuse. Here are some solutions we recommend:

Opt out

One of the most straightforward ways to protect your artwork from being used by AI without your permission is to opt out of platforms or services that might use your data for AI training.

Check the terms and conditions and settings of any platform you use and look for opt-out options specifically related to data collection and AI usage. By taking control of where and how your art is shared, you can significantly reduce the risk of unauthorized use.

At Linearity, we won't use your files for AI training unless you actively opt in within the app. Find out more in our T&Cs.

Add a digital signature

Digital signatures offer a discreet yet powerful way to safeguard your artwork. Acting as unique codes embedded within the file, these signatures serve as markers of authenticity and ownership.

Unlike visible watermarks, digital signatures are hidden and can't be easily removed. They use encryption to create a unique identifier linked to both the artwork and the artist. This not only proves ownership but also deters potential infringers from copying or distributing your work without permission. They’re like a robust layer of protection, ensuring that your creations remain yours and yours alone.

Use image cloaking

Image cloaking is a smart technique that subtly tweaks the pixels of your artwork, making it hard for AI algorithms to learn from or replicate it, all while keeping these changes invisible to the human eye. In essence, it confuses AI systems, preventing them from accurately analyzing or copying your work.

This method is particularly useful for artists who frequently share their creations on social media and other online platforms. It provides peace of mind, knowing that your art is protected from unauthorized AI use.

There are several tools available to help you cloak your images before sharing them online. We recommend Glaze and Nightshade from the University of Chicago's Glaze Project.

Glaze works by subtly tweaking the pixels of your images in a way that's invisible to the human eye but confuses AI algorithms. This means AI can't easily analyze or copy your work, making Glaze perfect for artists who want to share their creations online without worrying about them being used for AI training.

Nightshade, on the other hand, adds hidden tags to your images using the Pytorch framework. These tags aren't noticeable to people, but they mess with AI systems trying to learn from your artwork. Nightshade also helps track and verify your work, ensuring any unauthorized changes are detectable.

Together, Glaze and Nightshade offer solid protection, keeping your art safe from AI misuse.

Adjust your social media settings

Another crucial step in protecting your artwork from AI exploitation is adjusting your social media settings to enhance privacy and control over your content.

Here's how you can do it:

1. Privacy settings: Change your social media settings to control who sees your posts and data. Make sure only people you trust can see your artwork.
2. Data sharing opt-out: Check if you can opt out of sharing your data for AI training. This stops your artwork from being used without your permission.
3. Restricted access: Share lower-resolution images to make it harder for AI to copy your work.
4. Be mindful of T&Cs: Always make sure to read the terms of service for each social media platform. Look for sections about data use, copyright, and AI training, and choose carefully where you want to share your work.

In this digital age, where the boundaries between art and technology blur, it's more important than ever for artists and designers to stay informed and empowered in the face of emerging challenges. By staying embracing these protective measures, you can continue to create and share your work with confidence.