In a bid to safeguard their artistic creations from the unauthorised use by AI image generators, artists are now able to deploy a tool called Nightshade which employs a technique called ‘data poisoning.”

The tool was developed by a team of computer scientists at the University of Chicago, and it works by subtly altering the pixels in the images of art pieces, rendering them unchanged to the human eye but causing chaos for computer vision. 

“Used responsibly, Nightshade can help deter model trainers who disregard copyrights, opt-out lists and do-not-scrape/robots.txt directives,” the Nightshade team said in a blog post. 

The more “poisoned” images introduced into the training data, the greater the disruption, potentially forcing rogue tech companies to respect copyright laws. 

“It does not rely on the kindness of model trainers, but instead associates a small incremental price on each piece of data scraped and trained without authorisation,” said the team. 

However, the tool has sparked concerns about its potential misuse, as intentionally “poisoning” the training data could be employed to purposely disrupt the services of AI generators. 

While Nightshade has a noble goal of protecting artists’ intellectual property, it raises ethical questions about the responsible use of such tools.

“Nightshade’s goal is not to break models, but to increase the cost of training on unlicensed data, such that licensing images from their creators becomes a viable alternative,” said the researchers in response to this critique. 

Recommended reading

• Vodafone and Microsoft Sign 10-year AI and Cloud Partnership
• CMA Seeks Views on Microsoft Partnership With OpenAI
• OpenAI Claims AI Training “Impossible” Without Copyrighted Material

Some of the research team behind Nightshade were also behind Glaze, which added subtle changes to artwork to interfere with the ability of AI models to read the image data. 

The researchers said Glaze is a defensive tool to prevent style mimicry, while Nightshade is designed to be an “offensive tool” that disrupts AI training data while making minimal visible changes to the original image.

In a related development, Meta Platforms Inc. introduced a new AI image generator named Imagine with Meta in December. This generator, along with similar offerings from Microsoft Corp.-backed OpenAI, has raised concerns about the potential misuse of such technologies.

The responsible and ethical use of tools like Nightshade will likely remain a topic of debate as artists seek to assert control over the fate of their digital creations in an increasingly AI-driven world.