Login
An employee at Elon Musk’s artificial intelligence company xAI leaked a private key on GitHub that for the past two months could have allowed anyone to query private xAI large language models (LLMs) which appear to have been custom made for working with internal data from Musk’s companies, including SpaceX, Tesla and Twitter/X, KrebsOnSecurity has learned.
Image: Shutterstock, @sdx15.
Philippe Caturegli, “chief hacking officer” at the security consultancy Seralys, was the first to publicize the leak of credentials for an x.ai application programming interface (API) exposed in the GitHub code repository of a technical staff member at xAI.
Caturegli’s post on LinkedIn caught the attention of researchers at GitGuardian, a company that specializes in detecting and remediating exposed secrets in public and proprietary environments. GitGuardian’s systems constantly scan GitHub and other code repositories for exposed API keys, and fire off automated alerts to affected users.
GitGuardian’s Eric Fourrier told KrebsOnSecurity the exposed API key had access to several unreleased models of Grok, the AI chatbot developed by xAI. In total, GitGuardian found the key had access to at least 60 fine-tuned and private LLMs.
“The credentials can be used to access the X.ai API with the identity of the user,” GitGuardian wrote in an email explaining their findings to xAI. “The associated account not only has access to public Grok models (grok-2-1212, etc) but also to what appears to be unreleased (grok-2.5V), development (research-grok-2p5v-1018), and private models (tweet-rejector, grok-spacex-2024-11-04).”
Fourrier found GitGuardian had alerted the xAI employee about the exposed API key nearly two months ago — on March 2. But as of April 30, when GitGuardian directly alerted xAI’s security team to the exposure, the key was still valid and usable. xAI told GitGuardian to report the matter through its bug bounty program at HackerOne, but just a few hours later the repository containing the API key was removed from GitHub.
“It looks like some of these internal LLMs were fine-tuned on SpaceX data, and some were fine-tuned with Tesla data,” Fourrier said. “I definitely don’t think a Grok model that’s fine-tuned on SpaceX data is intended to be exposed publicly.”
xAI did not respond to a request for comment. Nor did the 28-year-old xAI technical staff member whose key was exposed.
Carole Winqwist, chief marketing officer at GitGuardian, said giving potentially hostile users free access to private LLMs is a recipe for disaster.
“If you’re an attacker and you have direct access to the model and the back end interface for things like Grok, it’s definitely something you can use for further attacking,” she said. “An attacker could it use for prompt injection, to tweak the (LLM) model to serve their purposes, or try to implant code into the supply chain.”
The inadvertent exposure of internal LLMs for xAI comes as Musk’s so-called Department of Government Efficiency (DOGE) has been feeding sensitive government records into artificial intelligence tools. In February, The Washington Post reported DOGE officials were feeding data from across the Education Department into AI tools to probe the agency’s programs and spending.
The Post said DOGE plans to replicate this process across many departments and agencies, accessing the back-end software at different parts of the government and then using AI technology to extract and sift through information about spending on employees and programs.
“Feeding sensitive data into AI software puts it into the possession of a system’s operator, increasing the chances it will be leaked or swept up in cyberattacks,” Post reporters wrote.
Wired reported in March that DOGE has deployed a proprietary chatbot called GSAi to 1,500 federal workers at the General Services Administration, part of an effort to automate tasks previously done by humans as DOGE continues its purge of the federal workforce.
A Reuters report last month said Trump administration officials told some U.S. government employees that DOGE is using AI to surveil at least one federal agency’s communications for hostility to President Trump and his agenda. Reuters wrote that the DOGE team has heavily deployed Musk’s Grok AI chatbot as part of their work slashing the federal government, although Reuters said it could not establish exactly how Grok was being used.
Caturegli said while there is no indication that federal government or user data could be accessed through the exposed x.ai API key, these private models are likely trained on proprietary data and may unintentionally expose details related to internal development efforts at xAI, Twitter, or SpaceX.
“The fact that this key was publicly exposed for two months and granted access to internal models is concerning,” Caturegli said. “This kind of long-lived credential exposure highlights weak key management and insufficient internal monitoring, raising questions about safeguards around developer access and broader operational security.”
President Trump last week revoked security clearances for Chris Krebs, the former director of the Cybersecurity and Infrastructure Security Agency (CISA) who was fired by Trump after declaring the 2020 election the most secure in U.S. history. The White House memo, which also suspended clearances for other security professionals at Krebs’s employer SentinelOne, comes as CISA is facing huge funding and staffing cuts.
Chris Krebs. Image: Getty Images.
The extraordinary April 9 memo directs the attorney general to investigate Chris Krebs (no relation), calling him “a significant bad-faith actor who weaponized and abused his government authority.”
The memo said the inquiry will include “a comprehensive evaluation of all of CISA’s activities over the last 6 years and will identify any instances where Krebs’ or CISA’s conduct appears to be contrary to the administration’s commitment to free speech and ending federal censorship, including whether Krebs’ conduct was contrary to suitability standards for federal employees or involved the unauthorized dissemination of classified information.”
CISA was created in 2018 during Trump’s first term, with Krebs installed as its first director. In 2020, CISA launched Rumor Control, a website that sought to rebut disinformation swirling around the 2020 election.
That effort ran directly counter to Trump’s claims that he lost the election because it was somehow hacked and stolen. The Trump campaign and its supporters filed at least 62 lawsuits contesting the election, vote counting, and vote certification in nine states, and nearly all of those cases were dismissed or dropped for lack of evidence or standing.
When the Justice Department began prosecuting people who violently attacked the U.S. Capitol on January 6, 2021, President Trump and Republican leaders shifted the narrative, claiming that Trump lost the election because the previous administration had censored conservative voices on social media.
Incredibly, the president’s memo seeking to ostracize Krebs stands reality on its head, accusing Krebs of promoting the censorship of election information, “including known risks associated with certain voting practices.” Trump also alleged that Krebs “falsely and baselessly denied that the 2020 election was rigged and stolen, including by inappropriately and categorically dismissing widespread election malfeasance and serious vulnerabilities with voting machines” [emphasis added].
Krebs did not respond to a request for comment. SentinelOne issued a statement saying it would cooperate in any review of security clearances held by its personnel, which is currently fewer than 10 employees.
Krebs’s former agency is now facing steep budget and staff reductions. The Record reports that CISA is looking to remove some 1,300 people by cutting about half its full-time staff and another 40% of its contractors.
“The agency’s National Risk Management Center, which serves as a hub analyzing risks to cyber and critical infrastructure, is expected to see significant cuts, said two sources familiar with the plans,” The Record’s Suzanne Smalley wrote. “Some of the office’s systematic risk responsibilities will potentially be moved to the agency’s Cybersecurity Division, according to one of the sources.”
CNN reports the Trump administration is also advancing plans to strip civil service protections from 80% of the remaining CISA employees, potentially allowing them to be fired for political reasons.
The Electronic Frontier Foundation (EFF) urged professionals in the cybersecurity community to defend Krebs and SentinelOne, noting that other security companies and professionals could be the next victims of Trump’s efforts to politicize cybersecurity.
“The White House must not be given free reign to turn cybersecurity professionals into political scapegoats,” the EFF wrote. “It is critical that the cybersecurity community now join together to denounce this chilling attack on free speech and rally behind Krebs and SentinelOne rather than cowering because they fear they will be next.”
However, Reuters said it found little sign of industry support for Krebs or SentinelOne, and that many security professionals are concerned about potentially being targeted if they speak out.
“Reuters contacted 33 of the largest U.S. cybersecurity companies, including tech companies and professional services firms with large cybersecurity practices, and three industry groups, for comment on Trump’s action against SentinelOne,” wrote Raphael Satter and A.J. Vicens. “Only one offered comment on Trump’s action. The rest declined, did not respond or did not answer questions.”
On April 3, President Trump fired Gen. Timothy Haugh, the head of the National Security Agency (NSA) and the U.S. Cyber Command, as well as Haugh’s deputy, Wendy Noble. The president did so immediately after meeting in the Oval Office with far-right conspiracy theorist Laura Loomer, who reportedly urged their dismissal. Speaking to reporters on Air Force One after news of the firings broke, Trump questioned Haugh’s loyalty.
Gen. Timothy Haugh. Image: C-SPAN.
Virginia Senator Mark Warner, the top Democrat on the Senate Intelligence Committee, called it inexplicable that the administration would remove the senior leaders of NSA-CYBERCOM without cause or warning, and risk disrupting critical ongoing intelligence operations.
“It is astonishing, too, that President Trump would fire the nonpartisan, experienced leader of the National Security Agency while still failing to hold any member of his team accountable for leaking classified information on a commercial messaging app – even as he apparently takes staffing direction on national security from a discredited conspiracy theorist in the Oval Office,” Warner said in a statement.
On Feb. 28, The Record’s Martin Matishak cited three sources saying Defense Secretary Pete Hegseth ordered U.S. Cyber Command to stand down from all planning against Russia, including offensive digital actions. The following day, The Guardian reported that analysts at CISA were verbally informed that they were not to follow or report on Russian threats, even though this had previously been a main focus for the agency.
A follow-up story from The Washington Post cited officials saying Cyber Command had received an order to halt active operations against Russia, but that the pause was intended to last only as long as negotiations with Russia continue.
The Department of Defense responded on Twitter/X that Hegseth had “neither canceled nor delayed any cyber operations directed against malicious Russian targets and there has been no stand-down order whatsoever from that priority.”
But on March 19, Reuters reported several U.S. national security agencies have halted work on a coordinated effort to counter Russian sabotage, disinformation and cyberattacks.
“Regular meetings between the National Security Council and European national security officials have gone unscheduled, and the NSC has also stopped formally coordinating efforts across U.S. agencies, including with the FBI, the Department of Homeland Security and the State Department,” Reuters reported, citing current and former officials.
President’s Trump’s institution of 125% tariffs on goods from China has seen Beijing strike back with 84 percent tariffs on U.S. imports. Now, some security experts are warning that the trade war could spill over into a cyber conflict, given China’s successful efforts to burrow into America’s critical infrastructure networks.
Over the past year, a number of Chinese government-backed digital intrusions have come into focus, including a sprawling espionage campaign involving the compromise of at least nine U.S. telecommunications providers. Dubbed “Salt Typhoon” by Microsoft, these telecom intrusions were pervasive enough that CISA and the FBI in December 2024 warned Americans against communicating sensitive information over phone networks, urging people instead to use encrypted messaging apps (like Signal).
The other broad ranging China-backed campaign is known as “Volt Typhoon,” which CISA described as “state-sponsored cyber actors seeking to pre-position themselves on IT networks for disruptive or destructive cyberattacks against U.S. critical infrastructure in the event of a major crisis or conflict with the United States.”
Responsibility for determining the root causes of the Salt Typhoon security debacle fell to the Cyber Safety Review Board (CSRB), a nonpartisan government entity established in February 2022 with a mandate to investigate the security failures behind major cybersecurity events. But on his first full day back in the White House, President Trump dismissed all 15 CSRB advisory committee members — likely because those advisers included Chris Krebs.
Last week, Sen. Ron Wyden (D-Ore.) placed a hold on Trump’s nominee to lead CISA, saying the hold would continue unless the agency published a report on the telecom industry hacks, as promised.
“CISA’s multi-year cover up of the phone companies’ negligent cybersecurity has real consequences,” Wyden said in a statement. “Congress and the American people have a right to read this report.”
The Wall Street Journal reported last week Chinese officials acknowledged in a secret December meeting that Beijing was behind the widespread telecom industry compromises.
“The Chinese official’s remarks at the December meeting were indirect and somewhat ambiguous, but most of the American delegation in the room interpreted it as a tacit admission and a warning to the U.S. about Taiwan,” The Journal’s Dustin Volz wrote, citing a former U.S. official familiar with the meeting.
Meanwhile, China continues to take advantage of the mass firings of federal workers. On April 9, the National Counterintelligence and Security Center warned (PDF) that Chinese intelligence entities are pursuing an online effort to recruit recently laid-off U.S. employees.
“Foreign intelligence entities, particularly those in China, are targeting current and former U.S. government (USG) employees for recruitment by posing as consulting firms, corporate headhunters, think tanks, and other entities on social and professional networking sites,” the alert warns. “Their deceptive online job offers, and other virtual approaches, have become more sophisticated in targeting unwitting individuals with USG backgrounds seeking new employment.”
Image: Dni.gov
As Reuters notes, the FBI last month ended an effort to counter interference in U.S. elections by foreign adversaries including Russia, and put on leave staff working on the issue at the Department of Homeland Security.
Meanwhile, the U.S. Senate is now considering a House-passed bill dubbed the “Safeguard American Voter Eligibility (SAVE) Act,” which would order states to obtain proof of citizenship, such as a passport or a birth certificate, in person from those seeking to register to vote.
Critics say the SAVE Act could disenfranchise millions of voters and discourage eligible voters from registering to vote. What’s more, documented cases of voter fraud are few and far between, as is voting by non-citizens. Even the conservative Heritage Foundation acknowledges as much: An interactive “election fraud map” published by Heritage lists just 1,576 convictions or findings of voter fraud between 1982 and the present day.
Nevertheless, the GOP-led House passed the SAVE Act with the help of four Democrats. Its passage in the Senate will require support from at least seven Democrats, Newsweek writes.
In February, CISA cut roughly 130 employees, including its election security advisors. The agency also was forced to freeze all election security activities pending an internal review. The review was reportedly completed in March, but the Trump administration has said the findings would not be made public, and there is no indication of whether any cybersecurity support has been restored.
Many state leaders have voiced anxiety over the administration’s cuts to CISA programs that provide assistance and threat intelligence to election security efforts. Iowa Secretary of State Paul Pate last week told the PBS show Iowa Press he would not want to see those programs dissolve.
“If those (systems) were to go away, it would be pretty serious,” Pate said. “We do count on a lot those cyber protections.”
Pennsylvania’s Secretary of the Commonwealth Al Schmidt recently warned the CISA election security cuts would make elections less secure, and said no state on its own can replace federal election cybersecurity resources.
The Pennsylvania Capital-Star reports that several local election offices received bomb threats around the time polls closed on Nov. 5, and that in the week before the election a fake video showing mail-in ballots cast for Trump and Sen. Dave McCormick (R-Pa.) being destroyed and thrown away was linked to a Russian disinformation campaign.
“CISA was able to quickly identify not only that it was fraudulent, but also the source of it, so that we could share with our counties and we could share with the public so confidence in the election wasn’t undermined,” Schmidt said.
According to CNN, the administration’s actions have deeply alarmed state officials, who warn the next round of national elections will be seriously imperiled by the cuts. A bipartisan association representing 46 secretaries of state, and several individual top state election officials, have pressed the White House about how critical functions of protecting election security will perform going forward. However, CNN reports they have yet to receive clear answers.
Nevada and 18 other states are suing Trump over an executive order he issued on March 25 that asserts the executive branch has broad authority over state election procedures.
“None of the president’s powers allow him to change the rules of elections,” Nevada Secretary of State Cisco Aguilar wrote in an April 11 op-ed. “That is an intentional feature of our Constitution, which the Framers built in to ensure election integrity. Despite that, Trump is seeking to upend the voter registration process; impose arbitrary deadlines on vote counting; allow an unelected and unaccountable billionaire to invade state voter rolls; and withhold congressionally approved funding for election security.”
The order instructs the U.S. Election Assistance Commission to abruptly amend the voluntary federal guidelines for voting machines without going through the processes mandated by federal law. And it calls for allowing the administrator of the so-called Department of Government Efficiency (DOGE), along with DHS, to review state voter registration lists and other records to identify non-citizens.
The Atlantic’s Paul Rosenzweig notes that the chief executive of the country — whose unilateral authority the Founding Fathers most feared — has literally no role in the federal election system.
“Trump’s executive order on elections ignores that design entirely,” Rosenzweig wrote. “He is asserting an executive-branch role in governing the mechanics of a federal election that has never before been claimed by a president. The legal theory undergirding this assertion — that the president’s authority to enforce federal law enables him to control state election activity — is as capacious as it is frightening.”
FreshRSS