Star Trek Was Right about Prompt Injection, Sorta
A recent IEEE Spectrum article warns that prompt injection attacks can cause large language models to waste tokens by overthinking, potentially enabling a DDoS-style attack on LLM servers. While moder…