Musk’s team was caught using a customized version of the Grok chatbot, without approval, to sift through sensitive government data inside the Department of Homeland Security. They used a modified version in an attempt to hide the fact that they were doing it.
This raises serious questions:
– Are they allowing AI to store or copy personal data from federal systems?
– Is Grok being trained on Americans’ sensitive or confidential information?
– Could this give Musk access to nonpublic federal contracts that benefit Tesla or SpaceX?
– What agencies besides DHS were involved?
– Who approved this internally, if anyone?
– Has any of this been disclosed to Congress or the public?
And here are the risks if Grok wasn’t running in a tightly isolated, secure environment:
– Data leaks: AI systems can retain and regurgitate sensitive information unless strictly sandboxed.
– Unauthorized training: If Grok is learning from this data, it could be embedding federal records into a private product.
– Surveillance without oversight: Americans’ data may be analyzed or flagged with no legal safeguards.
– Unfair advantage: Musk could gain inside knowledge to benefit his companies or outpace other AI providers.
– National security exposure: DHS systems handle high-risk, confidential material, this isn’t just personal data, it’s critical infrastructure.
This isn’t a technical hiccup. Its unchecked AI use in government, especially by someone with financial and political stakes, puts democracy and privacy at risk.
Source