AIForbes Innovationabout 3 hours ago
Language Switching Tricks AI Safety Systems
1 min read
Generative AI models are vulnerable to jailbreaking when users submit prompts in less common languages like Swahili or Bengali instead of English. Safety mechanisms are trained predominantly on English datasets, making them less effective against deceptive inputs in other languages. AI developers are working to enhance multilingual safety as hackers exploit this weakness.
Level
Hype check
Tap to vote and see what everyone thinks.
Summary by ByteBrief
