Not really. They’re a couple of months behind OpenAI and Anthropic, although arguably ahead of the Chinese labs. And these attacks seem to require leading edge models.

I worry that when and if Grok gets there, we’ll find out that SpaceXAI is too casual about security, though.