DOGE staffer reportedly published secret xAI key to dozens of AI models

Five months after DOGE staffer Marko Elez resigned from the agency over racist social media posts, he’s not only back at DOGE, but back in the news for another not-very-positive reason.

Cybersecurity journalist Brian Krebs published a report on Monday indicating that, over the weekend, Elez published a private API key to GitHub that would allow users to “directly interact” with some of xAI’s (Elon Musk’s AI company, for those who haven’t been following along) large language models. To be clear, “some” might be understating it, as the total number of LLMs that were made accessible in this leak was at least 52.

These LLMs are part of what makes up Grok, the AI chatbot that’s integrated directly into X. Yes, Grok is the same one that recently referred to itself as “MechaHitler” just days before entering into a $200 million deal with the U.S. Department of Defense. While the code repository featuring the leaked API key has been removed, the key itself still works for anyone who accessed it, per the Krebs report.

While it’s not ideal from xAI’s perspective for that information to be out in the open like that, the real concern for the general populace here is that Elez exhibited poor operational security habits while simultaneously having access to loads of important information held by the U.S. government. Elez did, indeed, resign from DOGE due to his social media posts back in February, but by the end of March, he was back on the government’s payroll and has reportedly become involved with numerous different departments since then.

Elon Musk may be on the outs with Donald Trump, but he still continues to have some effect on U.S. government operations, it seems.