AI • Cybersecurity • AISafety
A handful of poisoned training documents can backdoor the world's largest AI models. Researchers...
A handful of poisoned training documents can backdoor the world's largest AI models. Researchers just proved it.
Felix Ghauri ·