New attack lets hackers plant hidden instructions in AI memory with a single prompt

A newly demonstrated attack technique allows hackers to plant hidden instructions inside an AI agent’s memory with a single prompt, enabling them to influenc...