CAN AI MODELS BE JAILBROKEN TO PHISH ELDERLY VICTIMS? AN END-TO-END EVALUATION

Summary from the TLDR Dev newsletter

AI can be used to scam elderly people. These journalists successfully jailbroke AI models like Meta and Gemini to generate phishing emails, which were then sent to elderly participants, resulting in 11% being phished. The journalists also investigated real-world AI scam factories in Southeast Asia and found that victims were using AI chatbots to conduct scams.