they do that when they're not hacked, too. makes it very tricky to detect
The hardware underpinning major AI models can be hacked so that chatbots provide faulty and dangerous answers to questions from users, according to new research shared exclusively with CNN. https://cnn.it/3W9TI6N