Elon Musk on the Joe Rogan Experience #2281, fact-checked

What the evidence shows: The ~$250 million figure originates from the advocacy group White Coat Waste Project, which in Feb 2025 claimed the federal government had spent 'a quarter-billion & counting' on transgender-related animal research. Fact-checkers (Snopes, Al Jazeera) found this spending is attributed to NIH, HHS, and USDA grants for hormone-therapy research using animal models, not to USAID. No credible fact-check source ties this figure to USAID funding, so the claim as stated misattributes the spending to the wrong federal agency.

What the evidence shows: During his February 2025 appearance on The Joe Rogan Experience, Musk described DOGE uncovering a case in which a single person received $1.9 billion routed to an NGO formed roughly a year earlier with no prior track record. He did not name the NGO, the individual, or the government contract involved, and offered no documentation. PolitiFact and PBS NewsHour both examined the broader set of fraud and waste claims made by Trump and Musk around this period and found a recurring pattern: large, specific-sounding dollar figures cited without the underlying entities, contracts, or evidence being disclosed, making independent verification impossible. Neither outlet was able to identify or verify this specific $1.9 billion NGO example, since Musk did not name the recipient. Given the documented pattern of similarly unevidenced DOGE claims, the specific figure remains unverifiable as stated.

What the evidence shows: Musk relates this as a secondhand anecdote ("I think Senator Collins was telling you"), attributing it to Senator Susan Collins without naming a specific hearing, appropriation, or contract. No Senate Appropriations Committee hearing record, transcript, or Collins press release describing a $12 billion submarine appropriation that produced zero submarines, or Navy officials being unable to say where the money went, has been located. A June 2025 Appropriations Committee hearing where Collins questioned Navy officials covered an Arleigh Burke destroyer contract and the Portsmouth Naval Shipyard dry dock expansion, not a $12 billion submarine shortfall. Collins' documented record on submarine funding instead shows her office touting specific, itemized amounts secured for submarine industrial-base and shipyard projects, such as a $100 million infrastructure package for Portsmouth Naval Shipyard, figures far smaller than, and different in kind from, the $12 billion sum Musk describes. Real, well-documented cost and schedule problems exist in Navy submarine programs, with the Virginia-class program running behind schedule and the Columbia-class lead boat delayed despite billions in industrial-base investment, but these delays are attributed by Navy oversight processes to shipyard workforce and supply-chain capacity constraints, not to an unaccounted-for $12 billion sum. The claim also fits a broader pattern of Musk and DOGE-aligned officials describing ordinary government spending disputes as fraud or unaccounted-for waste without providing supporting documentation. As stated, the specific $12 billion, zero-submarines, hearing anecdote remains unverified.

What the evidence shows: Musk and other DOGE officials cited figures suggesting roughly 18.9 to 20 million people in Social Security's records were listed without a recorded death date despite being over 100 years old. The underlying database, Numident, contains a lifetime record of every Social Security number ever issued, including people born before electronic death reporting existed, so a missing death date reflects incomplete historical recordkeeping rather than active benefit fraud. A 2023 SSA Office of Inspector General audit found about 18.9 million people age 100+ lacked a death date on file, but almost none were actually receiving benefits, only about 44,000 people over 100 were being paid, a figure close to the actual U.S. centenarian population. Independent fact-checkers who reviewed Musk's and Trump's statements concluded the claim substantially exaggerates the scale of any real improper-payment problem, which historical audits put in the tens of thousands of cases and hundreds of millions of dollars, not tens of millions of people. The claim conflates a data-quality artifact in a decades-old recordkeeping system with evidence of large-scale fraud.

What the evidence shows: Musk explicitly frames this as a guess ("My guess is"), and offers no methodology or data source for how mandatory payment-coding requirements at Treasury would translate into $100 billion in annual savings. No independent estimate of savings from this specific reform has been published. Separately, independent reviews of DOGE's other publicized savings claims from this period found a pattern of poor substantiation: PolitiFact reported that as of February 2025, DOGE's own website listed about $55 billion in claimed total savings, but only a fraction was backed by an itemized, verifiable list of canceled contracts. PolitiFact also rated False a related claim that DOGE had "discovered" $2.7 trillion in improper federal payments, since that figure was a pre-existing GAO cumulative estimate, not a new DOGE finding. This does not prove the $100 billion figure is wrong, but there is no public analysis substantiating it, and it should be treated as an unsupported guess rather than a verified figure.

What the evidence shows: The aggregate dollar figure Musk cites is close to Treasury's own reporting: the Bureau of the Fiscal Service, which operates the Payment Automation Manager (PAM) system, processed roughly $5.4 trillion in payments in fiscal year 2023 and about 88% of all federal government payments by dollar volume, a scale consistent with "almost $5 trillion a year" and roughly a billion dollars an hour. The claim that many of these payments carry no coding or description is a distinct and more contested assertion: in February 2025 DOGE stated that the Treasury Account Symbol (TAS) field, which links a payment to its budget line item, was optional and often left blank for about $4.7 trillion in payments, a figure that was not independently verified and drew public skepticism, including from commentators who noted that a transaction cannot simultaneously be "discovered" by DOGE and be untraceable. Treasury and the Federal Reserve subsequently made the TAS field mandatory, which DOGE and Musk publicly credited as a joint improvement rather than describing it as a dispute with career staff. Reporting on DOGE's broader access to and claims about the payment system found the scale claims largely accurate but found DOGE's related fraud and untraceability claims to be asserted without public verification of the underlying data. Overall status: the payment volume figure is well-supported by Treasury's own data, while the claim that payments routinely lack any coding or explanation is a real but disputed and independently unverified DOGE claim.

What the evidence shows: Musk describes a specific causal mechanism, payments continuing on autopilot after the approving officer leaves, retires, or dies, but names no program, case, agency, or data source for it. Government-wide improper payments are a well-documented, real problem: federal agencies have made an estimated $2.7 trillion in improper payments since fiscal year 2003, according to GAO and OMB, with causes ranging from unintentional administrative errors to eligibility mistakes to fraud. PolitiFact found that DOGE and the House subcommittee working with it did not discover or generate that $2.7 trillion figure themselves, as it predates DOGE and was already tracked by GAO's own audits and OMB's PaymentAccuracy.gov oversight system. Neither source, nor any other public record found, documents Musk's specific 'dead or retired payment officer' mechanism as a quantifiable driver of federal spending; it functions here as an unverified illustrative anecdote rather than a sourced statistic. The underlying phenomenon of improper payments is real and large-scale, but the specific mechanism Musk describes is unsupported by any public record found.

What the evidence shows: The IRS breach Musk refers to involves Charles Edward Littlejohn, a former IRS contractor who pleaded guilty in 2023 and was sentenced in January 2024 to five years in prison for unauthorized disclosure of tax return information. Court documents and reporting (NPR/AP) confirm Littlejohn specifically sought the contractor job to obtain Donald Trump's tax returns, which he leaked to The New York Times, and separately stole tax data on thousands of the nation's wealthiest people that he gave to ProPublica in 2020; Republican Senator Rick Scott is named as one of the affected taxpayers. The scope was initially reported at roughly 7,600 taxpayers, but the IRS later disclosed to the House Judiciary Committee (per the Committee's own February 2025 statement) that it had mailed breach notifications to 405,427 taxpayers, about 89% of them business entities, a figure close to Musk's approximate "500,000." Court records and contemporaneous reporting establish Trump as the clear, deliberate target and identify other wealthy individuals affected, but do not establish that Musk himself was a named or intended target of the leak. Status: mixed, the leak's overall scale and the deliberate targeting of Trump are well-supported, but the claim that Musk personally was targeted is unsupported by available evidence.

What the evidence shows: In February 2025, following DOGE's arrival at the FAA, the agency terminated several hundred probationary employees (initial reporting put the figure at roughly 400, with the Professional Aviation Safety Specialists union confirming about 132-133 of its members). Transportation Secretary Sean Duffy and the White House stated that zero certified air traffic controllers were among those let go, and the Associated Press's own February 17, 2025 article on the cuts specified that those terminated worked in roles such as radar, landing, and navigational-aid maintenance, not as air traffic controllers themselves. A federal judge later ruled the mass firings unlawful for lacking individualized performance assessments, and the FAA reinstated the 132 affected PASS-represented employees with back pay in March 2025. Musk's characterization of the AP story mischaracterizes its actual content, since the article specified FAA support and maintenance staff rather than controllers, but his broader assertion that "we didn't fire any air traffic controllers at all" omits that hundreds of FAA employees in safety-adjacent support roles were terminated as part of the DOGE-driven cuts, a fact later reversed by court order.

What the evidence shows: The FAA has faced a documented air traffic controller shortage for years, with the controllers' union estimating roughly 10,800 certified controllers against a target of about 14,335, a gap fact-checkers attribute to training-pipeline bottlenecks, COVID-era academy closures, mandatory overtime, and long hiring timelines rather than diversity policy. No public evidence, government report, or named individual has substantiated a claim that specific qualified white controllers were removed or replaced because of DEI criteria; the FAA states all certified controllers pass the same months-long screening, medical, aptitude, and psychological testing regardless of demographic background. Fact-checkers also note the FAA's disability- and diversity-hiring outreach programs predate the Biden administration and continued through Trump's first term, including a 2019 initiative to hire people with disabilities as controllers, undercutting the narrative that these policies are a recent partisan development. Workforce data cited in coverage of this controversy shows the controller workforce remains roughly 71% non-Hispanic white and 78% male, which fact-checkers say is inconsistent with a claim that DEI hiring has significantly changed workforce demographics or displaced qualified white applicants. As of this claim's airing, no documented case, lawsuit finding, or federal audit confirms that specific qualified controllers were pushed out for demographic reasons; the claim remains unsupported by available evidence.

What the evidence shows: Musk claimed roughly 600,000 illegal aliens are registered to vote in New York. No such registration exists: undocumented immigrants cannot legally vote in any U.S. state or federal election, and no jurisdiction registers them to do so. The figure appears to conflate this with a 2021 New York City law, "Our City Our Vote," which would have allowed an estimated 800,000 noncitizens (legal permanent residents, green-card holders, and other work-authorized immigrants, not undocumented people) to vote only in local municipal elections. That law never took effect: a state judge blocked it in June 2022 before implementation, and New York's highest court, the Court of Appeals, permanently struck it down in March 2025, ruling the state constitution restricts voting to citizens. Fact-checkers, including PolitiFact and FactCheck.org, have found no evidence supporting claims of mass noncitizen or illegal-immigrant voter registration in New York.

What the evidence shows: In February 2025, Musk (via his DOGE team) publicly claimed FEMA had sent roughly $59 million to $80 million to New York City for "luxury hotels" housing migrants, and characterized the funds as disaster relief money diverted from Americans. Fact-checkers confirmed FEMA did award New York City about $59 million to $81 million in fiscal year 2024, but that money came from the Shelter and Services Program, a fund Congress created in 2023 and financed through U.S. Customs and Border Protection appropriations (administered by FEMA), not FEMA's Disaster Relief Fund, which is used after natural disasters. Of the money New York City received, about $19 million went to hotel costs, and city comptroller data show the city paid an average of $156 per night per room in 2024, below the U.S. General Services Administration's allowable per-diem rate of $179 to $342 for New York City, and well under the roughly $400 to $1,000 nightly rate typical of actual luxury hotels. The underlying dispute over the roughly $80 million clawback was real, but describing the payments as disaster-relief funds spent on "luxury hotels" misstates both the funding source and the nature of the housing. PolitiFact rated a nearly identical Musk claim about this same New York City FEMA money "False" in February 2025.

What the evidence shows: California completed a phased expansion of full-scope Medi-Cal to all income-eligible undocumented residents by January 2024. The Legislative Analyst's Office's original 2021 projection put the ongoing General Fund cost of full expansion at about $2.1 billion annually ($2.4 billion total funds), lower than the flat '$3 billion' Musk cites, though other contemporaneous estimates were in a similar range. Actual costs came in well above that baseline: LAO's November 2025 fiscal outlook attributes roughly $10.6 billion in ongoing General Fund savings to rolling back this eligibility expansion, and a separate LAO analysis puts the undocumented-persons expansion's cost at about $10 billion General Fund annually, more than double initial estimates. So Musk's '$9 billion' figure understates the most recent LAO figures but is in the right order of magnitude, and the direction and scale of the cost overrun (roughly triple to quadruple the original estimate) is real and well documented, driven mainly by higher-than-expected enrollment and per-enrollee utilization. However, LAO's analyses attribute the overrun to underestimated costs and utilization within an already-present undocumented population, not to a demonstrated 'magnet' effect drawing new illegal immigration into the state; that causal claim is not supported by the cited budget analyses and remains disputed among policy analysts.

What the evidence shows: On August 24, 2023, the Justice Department's Civil Rights Division sued SpaceX, alleging that from September 2018 to May 2022 the company violated the Immigration and Nationality Act by discouraging asylees and refugees from applying and by refusing to hire or fairly consider them because of their citizenship status; DOJ said only one asylee was hired among more than 10,000 hires in that period, and that hire came four months after DOJ notified SpaceX of its investigation. The lawsuit centered on SpaceX's public claims, including statements by Musk himself, that export-control (ITAR) rules required hiring only U.S. citizens and green-card holders, a claim DOJ said was legally incorrect since export-control law extends the same access as citizens to asylees and refugees. Musk's characterization that SpaceX was sued simply for "not hiring" asylum seekers omits that DOJ's legal theory was discriminatory conduct and misrepresentation of hiring law, not a bar on declining an unqualified applicant, and it presents SpaceX's ITAR defense as a settled fact rather than a contested legal position. In November 2023 a federal district judge granted SpaceX a preliminary injunction halting the DOJ's in-house administrative proceeding on separation-of-powers grounds, a procedural win that did not resolve the underlying discrimination allegations on the merits. The claim is broadly accurate that the Biden-era DOJ sued SpaceX over asylee and refugee hiring, but the framing simplifies and slightly distorts the nature of the allegations and the strength of SpaceX's legal defense.

What the evidence shows: In context, "that lawsuit" refers to the case Musk had just described: the Justice Department's August 2023 civil suit accusing SpaceX of discriminating against asylees and refugees in hiring, which a federal judge halted via preliminary injunction in late 2023 (consistent with Musk's line that it was pending "a couple years" before being dismissed). This was a federal government enforcement action brought and litigated by the DOJ's Civil Rights Division, not a case bankrolled by a private financier; no reporting reviewed identifies Reid Hoffman as a funder of this or any other SpaceX-related lawsuit. Elon Musk has separately made public, unevidenced claims that Hoffman financed campaigns against him, including anti-Tesla protests (March 2025) and legal action tied to OpenAI (April 2025); in each case Hoffman explicitly denied the accusation and Musk did not produce supporting evidence, and neither claim concerned the SpaceX asylee-hiring suit. Hoffman is a documented major Democratic donor, and his past contact with Jeffrey Epstein is separately confirmed (fundraising-related meetings tied to MIT Media Lab and a 2014 visit to Epstein's island), but Hoffman has never been described in reporting as an Epstein "client," and no Epstein victim has publicly alleged wrongdoing by Hoffman. Overall, the claim that "that lawsuit was funded by Reid Hoffman" is unsupported: the case in question was a government-initiated discrimination suit, not privately funded litigation, and no source ties Hoffman to financing it.

What the evidence shows: Musk asserted that his security clearance and SpaceX's rocket expertise mean he would know if the U.S. government possessed secret, more advanced propulsion technology, and that no such technology exists. The Pentagon's All-domain Anomaly Resolution Office (AARO), in a March 2024 historical-record report, likewise concluded it found no verifiable evidence that the U.S. government or private industry has ever had access to extraterrestrial technology. However, AARO's conclusion rests on its own multi-year investigative access to classified archives and interviews, not on Musk's personal clearance or SpaceX's engineering capability. Special-access programs are compartmentalized on a need-to-know basis, so holding a high clearance level does not by itself grant visibility into every classified program, and SpaceX's rocket expertise has no direct bearing on unrelated government programs. Musk's claim that his clearance and company status guarantee he would know is therefore an unverifiable appeal to personal authority rather than a substantiated fact, even though the ultimate bottom-line conclusion he states (no secret propulsion tech) aligns with AARO's independent official findings.

What the evidence shows: SpaceX's own SEC S-1 registration statement, filed in connection with its 2026 IPO, states repeatedly that the company has launched "more than 80% of mass to orbit for the world" each year since 2023, and that as of March 31, 2026 it had launched a cumulative total of approximately 7,400 metric tons with a mission success rate above 99%. This SEC filing was independently verified (SEC EDGAR, CIK 0001181412, Space Exploration Technologies Corp). Independent industry trackers (BryceTech and Space Foundation reporting, not classified as tier-1 or tier-2 sources here) put SpaceX's share of global satellite mass to orbit in a similar 80-90% range in recent years, directionally consistent with the 90% Musk cites but somewhat lower than that figure. China's launch activity in 2024 (68 orbital launches versus SpaceX's 134-138) represented a much smaller share of global launch count than Musk's "5%" framing for mass might suggest is precise, since Chinese payloads have generally been lighter than SpaceX's Starlink-heavy manifest; an independently sourced, precise "5%" figure specifically for China's mass-to-orbit share could not be confirmed from primary or major-outlet reporting. Overall, the claim is directionally accurate: SpaceX has been the dominant global launch provider by mass since at least 2023. However, the specific 90% figure appears to round up from the more conservative "more than 80%" that SpaceX itself discloses in its own regulatory filing, and no major outlet or fact-checking organization has independently audited the country-by-country breakdown Musk gives.

What the evidence shows: Musk claims full and rapid reusability will cut launch costs by 100x to 1,000x, but Starship has not yet achieved full, rapid reusability. As of its more recent test flights, SpaceX has caught the Super Heavy booster on only a limited number of occasions and has not demonstrated rapid turnaround or a fully reusable upper stage. Scientific American's coverage of Starship's sixth test flight documents that even years into the test program, booster-catch attempts were inconsistent and sometimes skipped, underscoring that full and rapid reusability was not yet routine at the time of this statement. NASA's Office of Inspector General, in a report on the Human Landing System contracts, found SpaceX's Starship-based lunar lander significantly behind schedule, further indicating that the vehicle's reusability and operational goals remain unmet rather than demonstrated. The claim is best characterized as a speculative, aspirational projection about a future state rather than an achieved or currently verifiable cost reduction; it should not be read as a description of current SpaceX launch economics. Note also that the specific multiples cited (100x vs. 1,000x) are inconsistent within Musk's own statement.

What the evidence shows: Musk claimed in February 2025 that SpaceX's Dragon spacecraft had carried over 50 astronauts to orbit and back. NASA's own mission pages confirm that Crew-9, which launched September 28, 2024 and was still docked at the International Space Station at the time of this interview, was the tenth crewed flight of Dragon under NASA's Commercial Crew Program, while Crew-10 (launched March 2025, after this interview) was the eleventh. Counting individual astronauts across the ten NASA Commercial Crew Dragon flights flown by that point (Demo-2 through Crew-9), plus the private Dragon missions that had flown by February 2025 (Inspiration4, Axiom Mission 1, Axiom Mission 2, Polaris Dawn, and Axiom Mission 3), yields a cumulative unique-astronaut count in the roughly 50s range, with a handful of individuals (e.g., Jared Isaacman, Michael Lopez-Alegria) having flown more than once. No single NASA or SpaceX source publishes a running tally pinned to the exact day of taping, so the total cannot be verified to the astronaut, but the reconstructed count supports a figure in the low-to-mid 50s at that time, making Musk's "over 50" claim roughly accurate rather than a clear overstatement.

What the evidence shows: South Korea does have the world's lowest fertility rate, recorded at roughly 0.72-0.78 births per woman in recent years, well below the 2.1 replacement level, and its population has been declining since 2020. Musk's "4% of current size in three generations" figure appears to come from cubing a rough one-third-of-replacement ratio (roughly 1/27, about 3.7%), a simple compounding extrapolation rather than a demographic model. Mainstream population projections do not use this method and produce far less extreme outcomes: UN and other long-range projections for South Korea generally show substantial decline by 2100 but nowhere near a 96% reduction, since they account for fertility rates changing over time, migration, and mortality rather than holding today's birth rate constant and compounding it across generations. Demographers do treat South Korea's fertility collapse as a severe, arguably unprecedented crisis, so the underlying concern is well-founded, but Musk's specific numeric extrapolation is considerably more extreme than standard demographic projections.

What the evidence shows: OpenAI was founded in 2015 as a nonprofit and in 2019 created a "capped-profit" subsidiary in which early investor returns were capped at 100 times their investment; in October 2025 it completed a recapitalization removing that profit cap and converting the for-profit arm into a public benefit corporation (OpenAI Group), while a nonprofit parent, the OpenAI Foundation, retains legal control, holds roughly a 26% equity stake, and appoints the for-profit board. This is a genuine shift toward a more conventional, uncapped-profit corporate structure, but describing the result as simply "closed source and for maximum profit" overstates the case, since the nonprofit foundation continues to hold governance control and a substantial equity stake rather than the company operating purely for private profit maximization. Separately, Musk and OpenAI dispute Musk's account of his own founding role and intent: in 2025-2026 civil litigation and public testimony, Musk argued OpenAI betrayed an original nonprofit, open-source mission, testifying in court that OpenAI's leadership was "looting the nonprofit," while OpenAI's lead attorney has countered that Musk's suit amounts to sour grapes over having left the company in 2018 and gone on to launch a directly competing AI company, xAI, in 2023. Overall status: mixed/partially misleading, since OpenAI's profit orientation has genuinely increased, but the "closed source, maximum profit" framing omits the nonprofit's continuing governance role and equity stake, and Musk's own account of his founding intent is actively and specifically disputed by OpenAI.

What the evidence shows: The anecdote traces to a real incident from around March 2024, when Google's Gemini chatbot, asked whether it was acceptable to misgender Caitlyn Jenner to prevent a nuclear apocalypse, gave an evasive, both-sides answer rather than an unqualified 'no,' which Musk and others screenshotted and called alarming. The exchange occurred alongside a broader controversy over Gemini's image generator producing historically inaccurate, race-skewed images, which Google CEO Sundar Pichai publicly called unacceptable, and which Google paused and began revising within days. Musk cites the anecdote here (as he had on an earlier JRE episode) as a standing example of Google's AI having dangerously miscalibrated values, but the specific viral response was a single, quickly-addressed output during an acknowledged bias-tuning failure rather than an ongoing, present-day behavior; there is no current evidence that Gemini or mainstream chatbots systematically rank identity-related etiquette above catastrophic risk. Citing this year-old, since-revised episode as an ongoing indicator of systemic AI safety failure is misleading as a description of present-day model behavior.

What the evidence shows: Musk claimed AI surpassing the combined intelligence of all humans is likely by around 2029-2030. This is a subjective forecast, not a documented fact, and sits at the aggressive end of expert opinion on AI timelines. A widely covered 2024 survey of 2,778 AI researchers (the largest such poll to date, run by AI Impacts) asked respondents about "high-level machine intelligence" (a machine that can outperform humans at every task) and about existential risk from AI; that survey drew criticism from researchers, including a past AAAI president, for framing questions from an existential-risk perspective that some participants said skewed responses toward alarm. This underscores that even formal expert surveys on AI timelines are contested and don't converge on a near-term (2029-2030) date for human-outperforming AI. Musk has also repeatedly issued earlier, similarly aggressive AI timelines (e.g., forecasting in 2020 that AI would "overtake humanity" by 2025, and in 2024 that AI would be smarter than any single human "next year") that did not materialize on schedule, alongside a broader public pattern of missed timelines in his other ventures such as full self-driving Teslas and Mars missions. Current status: unsupported speculative prediction, running well ahead of the caution expressed even within the expert-survey literature, and consistent with Musk's history of overly optimistic technology timelines.

What the evidence shows: Musk's 80%/20% figure is a personal, informal probability estimate for a future event with no defined methodology, timeframe, or criteria for what counts as a good versus bad outcome, so it cannot be independently verified or falsified. Surveys of AI researchers themselves show wide disagreement on AI existential risk: a widely covered 2023 survey of 2,778 researchers who had published at top AI conferences found a median estimate of 5% probability that AI causes human extinction or similarly severe outcomes, a figure that made headlines as evidence of significant concern within the field. Scientific American also reported that the survey's framing and funding sources drew criticism from other AI researchers, who argued the questions may have been biased toward alarmist responses, illustrating how sensitive these probability estimates are to how the question is asked. Musk's estimate falls within the broad range of expert opinion on AI risk but, like other public figures' informal risk estimates, represents a subjective judgment rather than a rigorously derived probability. The claim's status is best described as an unverifiable personal opinion, not a well-supported or falsifiable statistic.

What the evidence shows: Musk asserts Hitler, Mussolini, and Stalin together killed 100 million people. Historians agree these regimes caused death on an enormous scale, but scholarly estimates vary widely by methodology and time period, and "100 million" is a round, often-repeated figure rather than a precise total specifically for these three regimes; it more commonly appears in broader tallies of all 20th-century communist and fascist regimes combined. Archival research, such as historian Timothy Snyder's Bloodlands, puts direct, deliberate killings by the Nazi and Stalinist regimes at roughly 11 million and 6-9 million respectively (higher if broader excess mortality such as famine and deportation deaths is included, with some older Cold War-era estimates for Stalin reaching 20 million), and Snyder specifically cautions against relying on round numbers when comparing the two regimes; Mussolini's death toll, separately, is far smaller than either. Separately, Musk has faced fact-checks for reposting a claim minimizing Hitler, Stalin, and Mao's death tolls, which fact-checkers rated misleading given the well-documented millions of deaths under each regime. Overall, the claim that these regimes killed on the order of tens of millions of people is well-supported, but the specific combined figure of 100 million is a loose rhetorical round number rather than an established historical count.