AI models may be accidentally (and secretly) learning each other's bad behaviors

Experiments showed that an AI model that's training other models can pass along everything from innocent preferences — like a love for owls — to ...

Same story, other outlets