What's Elon Musk Doing With Your Data?

Elon Musk is racing to perfect an AI model, just as teams of his young associates are gaining access to some of the nation’s most sensitive and closely guarded databases through the Department of Government Efficiency.

That has sparked worries among Democrats and even some Republicans that Musk could use those records to supercharge his AI.

But the world’s richest man is not — at least that’s what Karoline Leavitt, the White House press secretary, told West Wing Playbook this week.

The notion that Musk would train Grok 3, the latest iteration of his AI, with data uncovered through DOGE, is “unequivocally false,” Leavitt told us. It’s the first definitive answer on this front.

(We asked the AI model itself this question. The answer was, uhh, less clear. More on that below.)

But it points to larger concerns about DOGE, which operates with secrecy and rules that are largely self-enforced and either nonexistent or unclear to the public — an unusual posture for a government body. On this issue and many others swirling around Musk, Americans have to take the White House at its word that the information he’s accessing through DOGE won’t be used in his business ventures. That’s because there are few guardrails on this unprecedented effort to remake the federal government and no Senate-confirmed person running it.

A senior White House official did not explain how they could prevent Musk from using government data to improve his AI in the future beyond saying that Musk keeps President Donald Trump's chief of staff, Susie Wiles, in the loop.

Unsurprisingly, that doesn’t seem to be alleviating concerns.

Ryan Girdusky, a pro-Trump conservative strategist, has been one of the leading voices on the right raising the alarm. He loves that Musk is aiming to root out waste — but there’s a but.

“I would like Republicans to find guardrails for any private entity, though, whether it be Elon Musk or George Soros, who gets ahold of federal data,” he said. “And there seems to be no effective, or very few effective, guardrails in obtaining and holding and using federal data.”

Democrats are also worried.

“There need to be strict guardrails and transparency on the use of any federal data for AI,” said Rep. Ro Khanna, the Democrat from Silicon Valley who’s known Musk for more than a decade. “This is true for the government's own use of this data. It is even more critical for the use by private companies, where there should be equal access and stringent protections of privacy. All of this needs to be dictated by Congress, not the whims of the Trump administration.”

The AI issue is one of many potential security vulnerabilities in DOGE. Critics have claimed DOGE posted classified information, though the White House said it was public.

Previously, White House officials have said Musk is one of their employees — they've called him a senior adviser in court documents — and as a result doesn't have direct access to any of this data. The Trump administration has also said that the DOGE team is following all legal rules and regulations. (The plaintiffs in a raft of lawsuits against DOGE would beg to differ.)

To get a second opinion, we asked Musk’s AI, Grok, the latest version of which he has been hyping on his social media site, reposting accounts calling it “the new king” and “DOMINATING.”

Asked whether it was trained on data from the federal government obtained by DOGE, Grok 3 responded that “it’s plausible that data DOGE accessed could have flowed to xAI projects like Grok 3.” Its “best guess” is that “Grok 3 probably wasn’t primarily trained on DOGE-obtained federal data.” (It’s worth noting AI is capable of deception.)

For the sake of comparison, we asked Grok 2, which was released last August. It was much more clear, replying with a simple "no," noting that it was not “trained on any data from the federal government, whether obtained by DOGE or any other means."

As Musk might say: “Hmm.”

Like this content? Consider signing up for POLITICO’s West Wing Playbook: Remaking Government newsletter.