New Tool Allows Artists to "Poison" Digital Art to Thwart AI Systems

Artists Fight Back

Researchers at the University of Chicago have developed a groundbreaking tool that empowers artists to protect their digital artwork from being used to train artificial intelligence (AI) systems. Dubbed "Nightshade," this tool allows artists to deliberately contaminate datasets used for AI training, making it harder for AI systems to accurately interpret and generate outputs.

Tricking the AI

Nightshade works by modifying the pixels of a digital image to deceive AI systems. For example, an image of a cat can be altered to make an AI system mistakenly identify it as a dog, and vice versa. By introducing these "poisoned" images into the training data, the AI's ability to produce accurate and meaningful results is compromised.

A Potential Vulnerability

Vitaly Shmatikov, a professor at Cornell University, warns that there are currently no robust defenses against these types of attacks. This implies that even sophisticated AI models, like OpenAI's ChatGPT, could be vulnerable to such manipulation.

Expanding Artist Protection

Professor Ben Zhao and his team at the University of Chicago, the masterminds behind Nightshade, have been working on artist protection software called Glaze. Nightshade is an extension of Glaze, which allows artists to obfuscate the style of their artwork. For instance, a charcoal portrait can be transformed to appear as modern art to an AI system.

Coming Soon to Glaze

Nightshade will eventually be integrated into Glaze, making it a comprehensive tool for artists to safeguard their work. Glaze is currently available for free on the University of Chicago's website, offering artists the means to protect their digital creations in today's competitive digital world.