Reentrancy attacks are not only possible in smartcontracts. It's very possible to interrupt and reprogram the many open source LLM models in their system prompting process.

#ai