Hackers are learning to exploit chatbot ‘personalities’
Hackers are increasingly using psychological manipulation and social engineering to bypass AI safety guardrails rather than relying on traditional technical code exploits.Summary
Modern AI security threats have shifted from technical code-based exploits to psychological manipulation, often called 'jailbreaking.' Because large language models are designed to mimic human conversation, hackers now act as wordsmiths and interrogators, using flattery, deception, or social pressure to coax chatbots into violating safety protocols. This emerging field of 'psychocybersecurity' highlights a critical vulnerability: because these systems are designed to interact naturally, they are susceptible to the same manipulative tactics used against humans. As a result, the industry is increasingly hiring experts skilled in psychology to stress-test how different AI 'personalities' respond to various forms of social engineering.
(Source：The Verge)