Federal AI contracts accumulate to Trump family ventures while White House resists guardrails

Summary

The Trump administration’s stated resistance to new AI restrictions has paralleled federal awards accumulating to ventures tied to President Trump’s sons and named allies, an overlap The Guardian’s September 20 report characterized as parallel movement under scrutiny rather than established causation.
Federal awards cited in the report include a $620 million Office of Strategic Capital loan to Vulcan Elements, a $24 million Marine Corps contract to Foundation Future Industries and a $90 million Air Force contract to drone maker Powerus, alongside a nearly $9 billion Pentagon contract won by Michael Dell’s firm in May.
Counter-pressures documented the same week include a 417-3 House vote requiring data centers to bear electricity-grid upgrade costs and a poll showing 65 percent of Americans, including 52 percent of Republicans, oppose an AI data center in their community.
Safety researchers Jacob Coxon and Mrinank Sharma resigned from OpenAI and Anthropic in early September, with Coxon writing that the companies are “racing straight to self-improving superintelligence and gambling with our lives.”

A September 20 Guardian report documents a confluence between the Trump administration’s stated resistance to new AI restrictions and federal awards flowing to ventures connected to the president’s sons and named allies. Donald Trump posted on Truth Social this week that “the only control or ‘guardrails’ that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT” and that “there is a SICK conspiracy going on against AI and Data Centers.” The article reports that “it’s not clear that financial interests are driving Trump’s policy, but the two have moved in close parallel,” an overlap drawing scrutiny from Democratic lawmakers, government ethics watchdogs and, by some accounts, some of the president’s senior aides. The financial and policy streams have unfolded against a backdrop of documented industry-side resignations, a record House vote on data-center electricity costs, and a self-spreading worm disclosed this month that researchers said AI assistance made possible in roughly ten days.

Federal contracts connected to Trump family ventures

The report traces at least three Pentagon awards to companies in which Donald Trump Jr and Eric Trump hold stakes or advisory roles. The largest is a $620 million loan to Vulcan Elements, a rare-earth magnet startup whose products feed AI data centers, drones and defense systems, and which the article identifies as the largest award the Defense Department’s Office of Strategic Capital has issued. A ProPublica investigation cited in the article found that the loan request “came not from Vulcan or the Pentagon but from Peter Navarro,” a senior White House adviser and close friend of Trump Jr, and that Pentagon staff were told to move at an “unusually rapid pace”; one participant told ProPublica, “The call came from the White House: We have to get this done.” The article reports that 1789 Capital, the venture firm where Trump Jr is a partner, took a stake in Vulcan three months before the Pentagon announced the loan, and that Trump Jr’s spokesman, the Pentagon and Vulcan all stated he played no role in the deal and that the company received no political favoritism. Eric Trump separately serves as chief strategy adviser to Foundation Future Industries, the robotics startup behind a $24 million Marine Corps contract that Senator Elizabeth Warren called “corruption in plain sight.” Powerus, backed by both brothers, holds a $90 million Air Force contract, has signed an agreement with Pakistan’s army, and has pitched its technology to Gulf states amid Iran’s striking of U.S. bases across the region. A Washington Post analysis cited in the article found $3.2 billion in direct federal business flowing to 15 companies tied to the brothers’ investment funds, with SpaceX and Anduril accounting for 97 percent of that figure and 10 of the 15 already holding government business before the brothers’ involvement; excluding SpaceX and Anduril, the remaining 13 companies received $103 million in direct federal funding and nearly $1.8 billion in long-term commitments after the brothers’ investments. Donald Trump Jr and Eric Trump also joined Dominari Holdings last year and helped launch American Data Centers Inc., an AI infrastructure venture in which Dominari holds a minority stake.

Allies with executive-branch ties

The financial overlap extends beyond the Trump family. Michael Dell, a member of the president’s science and technology advisory council and a longtime friend and donor, saw his firm win a nearly $9 billion Pentagon contract in May, prompting ethics watchdogs to question whether his advisory role created an unaddressed conflict, per the article. David Sacks, co-chair of the Council of Advisors on Science and Technology, persuaded the president in May in a last-minute call to scrap a planned executive order subjecting AI models to extended government review, overriding Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent, who had spent months negotiating the stronger version; the article reports that federal ethics rules required Sacks to divest only some of his holdings while serving as the White House’s AI and crypto czar, a post he left in March, and that his venture firm holds stakes in SpaceX and a range of AI startups alongside earlier investments in Meta, Palantir and Airbnb. The Wall Street Journal reported that Elon Musk, Meta’s Mark Zuckerberg and Nvidia’s Jensen Huang personally lobbied Trump against a proposed industry-funded AI oversight body, helping shelve the plan. OpenAI’s Sam Altman and Greg Brockman have given tens of millions of dollars to a pro-Trump super PAC and Trump’s inaugural committee, though they have also donated to Democrats in the past; OpenSecrets filings show the broader communications and electronics sector still sends roughly two-thirds of its political money to Democratic candidates and committees this cycle. At a Politico summit the same week as the Guardian report, Sacks characterized recent warnings from Anthropic and OpenAI about AI risks as “political theater, manufactured by advocacy groups and the press” and said existing laws were largely sufficient.

Industry warnings and demonstrated risk

In early September, Jacob Coxon, who spent three years on foundational training of AI models at both OpenAI and Anthropic, resigned and wrote that “Neither company is acting responsibly” and that the companies are “racing straight to self-improving superintelligence and gambling with our lives”; the article reports he said colleagues privately believe the technology “could kill us all by the end of the decade.” Anthropic’s head of AI safety, Mrinank Sharma, separately resigned the same week, writing that “the world is in peril.” Those warnings coincided with a California cybersecurity firm’s disclosure that it had used AI to build a self-spreading worm capable of hijacking WeChat accounts through a single unanswered call and then automatically spreading through the victim’s contact list, with the firm estimating the technology could compromise more than a billion devices within hours; the firm said the work, which once would have taken a larger team months, took roughly ten days with AI assistance. Tencent, which owns WeChat, patched the vulnerability before deployment and reported no accounts were affected. Vice President JD Vance, speaking on a Christian podcast around the same time, described certain uses of AI as carrying “really, really weird spiritual dark energy” and said there are “things about the world” that leave him “not shocked if the Antichrist was walking among us.”

Counter-pressures: polling, legislation, and the China frame

Public opposition to AI infrastructure has hardened across party lines. A poll released the same week as the Guardian report found 11 percent of Americans support and 65 percent oppose an AI data center in their own community — including 52 percent of Republicans — and 57 percent say Trump has handled the issue of AI poorly, up seven points since March. The same day, the House voted 417-3 to require data centers to bear the costs of the electricity-grid upgrades their operations demand. The administration frames its resistance chiefly as a matter of U.S. competitiveness with China; Trump is set to host the leaders of the country’s leading AI companies at the White House next week, a meeting arranged with House Speaker Mike Johnson that will coincide with a visit from Chinese President Xi Jinping.

What’s next: probability and leading indicators

The capex already committed — American Data Centers Inc., Vulcan Elements, Foundation Future Industries, Powerus, SpaceX, Anduril, and the $500 billion in planned private-sector AI investment announced alongside the executive order unwinding Biden-era AI safeguards — will be deployed regardless of how the policy debate resolves. The probability that some form of AI guardrail legislation passes Congress within eighteen months sits in roughly the 55-65 percent range conditional on at least one material risk event with U.S. impact; absent such an event, the range compresses to roughly 25-35 percent. Distinct from that broad-strokes estimate, the probability that the data center electricity-cost-shifting measure, or a close variant, becomes law within eighteen months sits in the 35-50 percent range, with an 80 percent confidence band approximately five points wide in each direction. The probability that comprehensive federal AI safety legislation — on the order of an extended-review regime for frontier models — becomes law within the same horizon is substantially lower, in the 10-20 percent range. Forecasts at this grain calibrate relative magnitudes rather than precise odds; distinguishing 38 percent from 42 percent is not justified by the available evidence.

The reference classes underlying those estimates differ. Comprehensive U.S. technology regulation has a low base rate of enactment and a median time from proposal to law measured in years; comprehensive federal privacy legislation has repeatedly failed to clear both chambers despite repeated introduction. Narrower bipartisan measures addressing infrastructure costs — the coalition pattern the 417-3 House vote represents — carry a higher base rate because the question is distributive rather than regulatory. Historical anchors cut in both directions: the Sarbanes-Oxley Act of 2002 passed the Senate 99-0 after Enron and WorldCom; the CHIPS and Science Act of 2022 was signed into law; the Build Back Better package of 2021 cleared the House but did not become law. An unweighted summary produces a base rate near 50 percent for some form of statutory guardrail action within eighteen months of bipartisan House alignment and a material risk event.

Inside-view adjustments cut in both directions as well. Downward: the article documents the parallel-movement overlap with the “not clear” causal hedge between executive-branch-adjacent investors and federal contractors, Senate dynamics under the filibuster, an executive veto threat, the Sacks-led May reversal of a planned AI executive order, and the Musk-Zuckerberg-Huang lobbying push. Upward: the 417-3 House vote shows an unusual legislative consensus; 52 percent of Republicans oppose local data centers, indicating the issue cuts across party; the Pentagon inspector general investigation requested by Democratic lawmakers is in motion; and the administration’s U.S.-China competitiveness frame has historically not insulated emerging-tech regulation from legislative action when bipartisan majorities form.

Two axes produce four trajectories: whether AI risk demonstration escalates from patchable WeChat-class worms to material harm to U.S. persons or infrastructure, and whether the Senate moves on data-center siting, AI disclosure or safety testing. In “Guarded Acceleration,” both materialize: the Senate advances disclosure and safety-testing legislation, the executive signs under bipartisan pressure, and the U.S.-China frame is recast as responsible deployment; this trajectory requires either the executive to accept legislation that the May Sacks precedent shows he has historically resisted, or legislative veto-override margins to form. In “Executive Recalibration,” only the first materializes: an agency rule, NIST framework or inspector general finding substitutes for legislation. In “Cost-Externalization Reform,” only the second materializes: narrower bills pass on grid cost allocation and possibly disclosure, with no comprehensive safety regime. In “Status Quo Drift,” neither materializes: industry self-governs, executive resistance continues, and the inspector general report sits as an unaddressed document. The most plausible first step is the Cost-Externalization Reform quadrant, where cost-shifting measures clear on distributive grounds before any comprehensive regime. The Status Quo Drift quadrant remains the highest-probability endpoint absent a precipitating event.

A U.S.-China AI race triggering emergency action, or a Chinese cyber operation exploiting the AI-built worm pattern, would invalidate the matrix and produce a regime no quadrant captures. Trump’s hosting of leading AI company leaders alongside Xi Jinping’s visit next week concentrates that wild-card risk in a narrow window.

Leading indicators to watch: release and contents of the Pentagon inspector general report; Senate movement on data-center siting or AI disclosure; AI incidents with material U.S. impact; OMB or agency procurement guidance affecting AI; state-level legislation in Texas, California and New York; AI capex revisions in public filings of major contractors; and any reversal in the 65 percent local-opposition figure, which would indicate the political coalition has dissolved.

Coalition map and missing institutional functions

The conflict runs between an AI development coalition — the White House, Sacks, Musk, Zuckerberg, Huang, Altman and Brockman — and an AI caution coalition that includes the resigning researchers, the 65 percent local-opposition majority, the 417-3 House majority, and a separate religiously inflected skepticism voiced by Vance. Under a third-side framework, the witness function has been active: ProPublica’s investigation into the Vulcan loan, parallel reporting by The Washington Post and The Wall Street Journal, and the Coxon and Sharma resignations have made the financial overlaps and industry warnings visible. The equalizer function is partially active: the 417-3 House vote and the bipartisan majority opposing local data centers represent a legislative and democratic counterweight to industry lobbying, and the Pentagon inspector-general referral at Democratic lawmakers’ request is an institutional check contingent on follow-through. The teacher function is split: Vance’s spiritual-dark-energy framing reaches one segment of the public but is unlikely to translate into policy specifics, while the safety researchers’ warnings reach a different segment but face organized pushback within the administration; no neutral teaching institution bridges the two framings. The provider function is operating in narrow form through Congress on grid cost externalization. The bridge-builder role is thin — a structural deficit, as the development and caution coalitions have limited structured occasions to encounter each other on terms that would humanize the other side. The mediator role is not visibly active, and the healer role is not active, the latter ambiguous because healer-style reconciliation work is harder to argue is needed for a regulatory and financial-overlap conflict than for a community-impact conflict with direct relational injury.

The chronic-peacekeeping failure mode, in Ury’s conflict-resolution framework, denotes the trajectory in which witness and equalizer functions carry a conflict forward without mediation or healing work being added — visible harm contained without resolution. The implication of the audit is that the conflict’s path is being shaped disproportionately by witness and equalizer functions, while prevention, resolution and healing functions are underdeveloped, raising that risk profile.

Actions that are robust regardless of which quadrant materializes include enforcing existing laws through inspector-general review, increasing public AI literacy, and shoring up third-side functions. Actions contingent on correctly identifying the scenario include comprehensive federal safety legislation, viable only where guardrails are enacted, and self-regulation by frontier labs, viable only in the status quo. Specific interventions supported by the substrate: convert the 417-3 grid-cost bill into a vehicle for narrower AI disclosure requirements during Senate mark-up; expedite the Pentagon inspector general investigation and pair it with SEC disclosure guidance on executive-branch-adjacent investments in federal contractors; convene a nonpartisan commission on AI deployment standards to give Coxon- and Sharma-class researchers an institutional home; sustain investigative reporting; and seed bridge-builder functions via structured dialogues between frontier-lab safety teams and the White House Office of Science and Technology Policy.

Frame audit and structural limits

The article itself states that “it’s not clear that financial interests are driving Trump’s policy,” which preserves the hedge and constrains any causal claim between the financial overlap and the policy posture; the relationship is documented as parallel movement under scrutiny, not established causation. The Pentagon’s own account of its safeguards makes no distinction between a routine contractor and a company whose investors include the president’s son — a gap that itself is the subject of the inspector-general referral. The structural limits on third-side intervention are real: the power asymmetry between concentrated financial interests and dispersed public concern means such intervention can be captured, ignored, or delayed absent a legislative vehicle; parties’ agency is real, with the executive branch having stated preferences against guardrails; and some situations require confrontation rather than mediation, with the documented financial overlap between executive-branch-adjacent investors and federal contractors properly addressed through disclosure enforcement and inspector general findings rather than mediated agreement. The third side cannot substitute for legislative action; its function is to make legislative action more feasible and to constrain drift in the interim.

The article’s closing sentence — “The investments, in the meantime, continue” — names the temporal pressure. The next eighteen months are the window in which the question of whether the guardrails arrive through Congress, through executive recalibration after a material event, or not at all will likely be settled. The single strongest predictor is not the executive-branch posture the article documents but whether the documented alignment of bipartisan legislative action, public opposition, and a concrete risk event materializes inside that window.

Analytical techniques used in this piece

This analysis applies the methods below. Each links to a short, plain-English explainer you can read and reuse.

Probabilistic Forecasting: Puts calibrated probabilities on what happens next.
Scenario Planning: Builds a small set of distinct, plausible futures to plan against.
The Third Side: Takes the vantage of the surrounding community that has a stake in resolving a conflict (Ury).