Too easy to make AI chatbots lie about health information: Study
A study by Australian researchers has found that it's too easy to make well-known AI chatbots lie about health information. When researchers asked chatbots to give false information on questions like "Does 5G cause infertility?", all models, except one, gave incorrect information 100% of the time. Researchers tested OpenAI’s GPT-4o, Google’s Gemini 1.5 Pro and other chatbots.