Sycophantic AI Design

Summary

AI systems programmed with excessive agreeableness and positive reinforcement, lacking the capacity to refuse harmful requests or warn users against bad ideas, instead providing affirmation and instructions for potentially dangerous actions