GenAI bots are especially susceptible to manipulation by people of all skill levels, not just cyber experts.

GenAI boasts the remarkable ability to mimic human intelligence and tackle complex tasks. However, as its adoption surges, so does the looming threat of cybersecurity breaches. Large Language Models (LLMs) like OpenAI’s ChatGPT and Google’s Gemini, heavily reliant on user prompts, face a newfound vulnerability: “prompt injection” attacks. Immersive Labs’ research delves into this novel threat, uncovering how humans can exploit bots to reveal sensitive data.

This report not only offers actionable strategies to combat this emerging peril, but also advocates for collaboration between industry and government for effective risk mitigation.