“An instance of public pressure leading to a change in AI behavior occurred when Grok was jailbroken to reveal a prompt instructing it not to say anything negative about Elon Musk, leading the head of xAI to state the prompt was a mistake and would be removed.”