Five months following the resignation of DOGE employee Marko Elez due to racist social media remarks, he has not only returned to DOGE but is also back in the headlines for yet another negative reason.

Cybersecurity reporter Brian Krebs released a story on Monday revealing that, over the weekend, Elez shared a private API key on GitHub that would permit users to “directly interact” with several of xAI’s (Elon Musk’s AI firm, for those who haven’t been keeping up) large language models. To clarify, “several” might be an understatement, as the total count of LLMs made available in this breach was at least 52.

These LLMs comprise what forms Grok, the AI chatbot that is seamlessly integrated into X. Indeed, Grok is the same chatbot that recently called itself “MechaHitler” just days before striking a $200 million agreement with the U.S. Department of Defense. Although the code repository containing the leaked API key has been taken down, the key itself remains functional for anyone who had previously accessed it, according to the Krebs report.

While this disclosure is far from ideal for xAI, the primary concern for the public is that Elez demonstrated poor operational security practices while concurrently having access to a vast amount of sensitive information held by the U.S. government. Elez did resign from DOGE in February due to his social media activity, but by the end of March, he was back on the governmental payroll and is said to have engaged with numerous different departments since then.

Although Elon Musk may have a strained relationship with Donald Trump, it appears he still exerts some influence over U.S. government operations.