Hackers could one day use novel visual techniques to manipulate what AI sees - RisingAttacK impacts 'most widely used AI computer vision systems'

Editor’s Note: As AI vision systems become more embedded in everything from self-driving cars to surveillance, vulnerabilities like this could have serious real-world consequences. If hackers exploit these flaws, they could manipulate AI in dangerous or deceptive ways—think traffic signs misread by autonomous vehicles or security cameras missing critical details. It’s a wake-up call for stronger defenses in AI systems we increasingly rely on.