Generative AI models are susceptible to a kind of cyberattack called “data poisoning,” whereby malicious actors intentionally manipulate known source material to change the model’s understanding of an issue. It’s like a high-tech version of giving a school rival a fake exam answer key.
Researchers say that concerns of data poisoning are mostly hypothetical at this point, but showed in a new report how Wikipedia entries could be edited at strategic times to ensure the incorrect information is captured by models scraping the online encyclopedia. It’s an early warning to AI companies and those who depend on it that attackers could soon find creative ways to target the most powerful models and exploit vulnerabilities.
Data poisoning isn’t all bad: Some copyright holders are using a form of data poisoning as a defensive mechanism to prevent AI models from gobbling up their creative works. One program called Nightshade was developed to distort an image when it’s ingested by a large language model.
Student protesters are set to take to the streets in Serbia this weekend in the first major demonstrations of the year against President Aleksandar Vučić, as frustration mounts over democratic backsliding and corruption.
Think you know what's going on around the world? Here's your chance to prove it.
GZERO Media is back on the podium at the 47th Annual Telly Awards, adding six more trophies to our shelf — including three in Gold! We’re so grateful to be recognized for our groundbreaking work in global analysis and… *checks notes*... geopolitical puppetry.
Ukraine is increasingly finding vulnerabilities in Russia’s defenses, piling further pressure on President Vladimir Putin.