This, and of course, this is the marketing people jumping up and down, saying, see, we have Mythos too. This incident implies these theoretical capabilities do apply in real-world settings. So, as I said, nice marketing for OpenAI, whose models have been seen as somewhat less capable than Anthropics, you know, since Mythos' marketing coup. This will, you know, tend to give more of the spotlight to OpenAI for a while. And that's a fair outcome, right? Because they really are. We know that these frontier models are really at near parity. Okay. So, next, we're going to look at the victim attack ease statement, meaning, you know, to see how Hugging Face views the event of having their security penetrated by OpenAI's road models.

You know, the thing that their agents discovered in order to get loose is being fixed. Fourth, we've brought Hugging Face into the trusted access program, meaning their trusted access program, and are supporting their teams in rapidly using our model's capabilities to improve their defenses. So, in other words, Hugging Face is saying, you know, WTF, we need to be safe against agents of this strength. Could you allow us to use yours as you have to make sure that we're secure? And so, they brought Hugging Face into OpenAI's trusted access program to have access to these new unrestrained and unreleased models. And finally, they said, we're improving and adding stronger protections around future training and evaluations.