On June 2, 2026, President Trump signed an executive order asking AI companies to share their most advanced frontier models with the federal government up to 30 days before public release. The administration frames this as voluntary cooperation, not regulation. Whether that distinction holds in practice is a different matter entirely.
The Thesis: No Rules Should Slow America Down
The order opens with familiar ground. The United States leads the world in AI because it refuses to strangle innovation with heavy-handed rules. The administration points to its rollback of Biden-era AI oversight as evidence of that commitment, and doubles down on accelerating AI adoption across both the public and private sectors.
The political signal is unmistakable: the global race is won by moving fast, not by regulating carefully. SpazioCrypto has covered this dynamic before, from agentic AI in banking to the broader ecosystem tracked in our AI section.
The Contradiction: Washington Wants to Look Under the Hood
Functionally, then comes the part that jars with the stated philosophy. The order asks companies to voluntarily share new frontier models with the federal government up to 30 days before launch. It also calls for identifying “trusted partners” among industry players, and directs the Committee on National Security Systems to strengthen the cyber defenses of critical networks within 30 days.
Voluntary sharing with the government before public release is, in practice, a form of state oversight dressed in softer language. The companies affected most are precisely those that carry the highest valuations and develop the most capable systems. When the White House makes a request of companies that depend on federal contracts and licenses, the line between voluntary and expected gets very thin.
Frontier AI Labs Affected (Private Valuation, bn USD)
Frontier AI Labs Affected (Private Valuation, bn USD)
Source: Reuters, Bloomberg · valuations at various dates (2026)
Source: Reuters, Bloomberg · valuations at various dates (2026)
The announcement came through official channels: → recent posts by @WhiteHouse on X.
What Does Trump’s AI Order Actually Require?
Three concrete provisions. First, voluntary sharing of frontier models with the federal government, with a window of up to 30 days before launch. Second, a process to identify “trusted partners” among AI companies, a mechanism that effectively determines who is inside the tent and who is not. Third, a directive to the Committee on National Security Systems to reinforce the cyber defenses of critical national security networks within 30 days. None of these are enforceable mandates with penalties attached. They are requests from the White House. For companies whose revenue depends on federal contracts and regulatory goodwill, the coercive weight is real even if the legal obligation is not. The cyber risk dimension is not theoretical: SpazioCrypto documented the first AI-generated zero-day exploit, recorded by Google.
Two Philosophies, One Technology
The contrast with Europe is sharp. Washington chooses the voluntary, national-security-framed path. Brussels chose binding rules calibrated to risk level, through the EU AI Act, with fines reaching 35 million euros or 7% of global turnover for serious violations.
While the U.S. deregulates and Italy deliberates, Spain transposed the AI Act into national law on May 26 through an organic law designed with SMEs in mind. The industrial race runs in parallel: from Nvidia’s $40 billion equity commitment to AI to model competition, illustrated when DeepSeek outperformed Bitcoin in a trading benchmark.
The primary source for the order’s full text is the White House presidential actions page.
The timing of the order is striking. It landed within days of Anthropic filing for its IPO and while OpenAI is preparing its own listing. The same labs now opening themselves to public-market scrutiny, with all the disclosure that entails, are simultaneously being asked to show their most advanced models to the government before anyone else sees them. Two very different forms of oversight, financial and state-level, are converging on the same handful of companies at the same moment. The question is no longer whether AI will be watched. It is who gets to do the watching.
