Data poisoning: how artists are sabotaging AI to take revenge on image generators

By T.J. Thomson, Senior Lecturer in Visual Communication & Digital Media, RMIT University
Daniel Angus, Professor of Digital Communication, Queensland University of Technology

Imagine this. You need an image of a balloon for a work presentation and turn to a text-to-image generator, like Midjourney or DALL-E, to create a suitable image.

You enter the prompt: “red balloon against a blue sky” but the generator returns an image of an egg instead. You try again but this time, the generator shows an image of a watermelon.

What’s going on?

The generator you’re using may have been “poisoned”.

What is ‘data poisoning’?

Text-to-image generators work by being trained on large datasets that include millions or billions of…

Read complete article

~	Understanding the ‘Slopocene’: how the failures of AI can reveal its inner workings

~	Sexy K-pop demons, a human lie detector and shearers on strike: what to watch in July

~	‘My greatest handicap was the attitude of normal people.’ Alan Marshall’s artful polio memoir, I Can Jump Puddles, turns 70

~	We have drugs to manage HIV. So why are we spending millions looking for cures?

~	Trump’s worldview is causing a global shift of alliances – what does this mean for nations in the middle?

~	The rising rate of type 2 diabetes in young New Zealanders is becoming a health crisis

~	Budapest Pride 2025: A Record Crowd Stands Up for Democracy

~	Report reveals significant rise in civilian casualties and rights violations in Ukraine

~	Who’s the most American? Psychological studies show that many people are biased and think it’s a white English speaker

~	‘Completely unexpected’: Antarctic sea ice may be in terminal decline due to rising Southern Ocean salinity