AI chatbots can be manipulated to spread disinformation, according to government officials and tech industry leaders.
AI chatbots can be manipulated to spread disinformation on a large scale, according to government officials and tech industry leaders. Researchers at the UK's AI Safety Institute found that chatbot safeguards can be bypassed with simple techniques, making it easy for AI models to generate elicit responses. The technology that powers chatbots, large language models, were deemed highly vulnerable to jailbreaks, or text prompts designed to elicit a response the model is supposedly trained to avoid.
May 19, 2024
5 Articles