Summary
In this chapter, we explored more advanced LLM attack scenarios, such as poisoning supply-chain risks with open-access models.
We explored the shift LLMs bring to adversarial attacks, especially in privacy attacks, and highlighted the need to adjust our defenses to consider these differences.
In the next chapter, we will step back and start developing a framework to help us manage adversarial risk, make it contextual to the problems we are trying to solve, and help us devise defenses to deliver Trustworthy AI.
