Tech industry warns against White House AI vetting regime proposal
Trump signed an executive order asking tech companies to voluntarily submit AI models for 30-day government review before release, rejecting earlier mandatory vetting proposals after industry pushback.
Objective Facts
President Trump signed an executive order on June 2, 2026, titled 'Promoting Advanced Artificial Intelligence Innovation and Security' directing federal agencies to establish a framework for secure deployment of frontier AI models, including a voluntary process by which developers would provide the government early access to models for up to 30 days before release to other trusted partners. A previous draft had called for a 90-day review window, but AI industry insiders pushed for something closer to two weeks, and Trump delayed signing the more demanding version in late May after industry pushback, including from former White House AI czar David Sacks. The order explicitly prohibits mandatory governmental licensing, preclearance, or permitting requirements for AI model development. Democratic Congressman Josh Gottheimer criticized the order's voluntary nature as lacking real teeth, arguing powerful AI frontier models exhibiting dangerous capabilities must be shared with the government before release, and that a purely voluntary framework means allowing AI to remain the Wild West. Major technology companies including Microsoft, Google, and OpenAI publicly welcomed the order as balancing innovation with security.
Left-Leaning Perspective
Progressive lawmakers like Congressman Josh Gottheimer led the Democratic critique of the executive order. Gottheimer stated on June 2, 2026 that while the order represents a step toward protecting America, its purely voluntary nature lacks real teeth—arguing that powerful AI frontier models exhibiting dangerous capabilities including cybersecurity risks and potential bioweapon creation must be shared with government before public release, and that a purely voluntary framework means allowing AI to remain the Wild West. The safety advocate community echoed these concerns about enforcement and scope. AI safety advocates argue it does not go far enough—lacking mandatory pre-deployment testing, binding safety standards, and independent oversight. Beyond these immediate reactions, critics questioned whether the timeline itself was adequate. Critics note Trump's revision compresses the review window into a single billing cycle of 30 days, barely enough time to schedule meetings let alone conduct meaningful analysis, with agencies that spent two years building 90-day review infrastructure now facing a 67 percent time reduction with no corresponding staff increase. The left's fundamental concern centered on the voluntary nature of compliance and lack of binding enforcement. Voices on the left warn that replacing government discretion on which CEO to brief with federal review staffed by intelligence community representatives but directly influenced by industry tech giants would not be better, given that nearly 80 percent of global AI computing power is privately owned and the majority of researchers capable of evaluating frontier systems work for the labs building them. This reflected skepticism that industry self-regulation, even with government observation, could be effective. What progressive coverage notably omitted was substantive engagement with arguments about competitive disadvantage against China, which Trump administration officials emphasized as the rationale for the lighter regulatory touch. The left focused almost exclusively on safety gaps and enforcement mechanisms rather than geostrategic considerations that shaped the final policy.
Right-Leaning Perspective
David Sacks, the venture capitalist and former White House AI czar who was instrumental in shaping the final order, provided the most detailed right-leaning defense of the policy. Sacks wrote that the voluntary framework applies only to models representing a meaningful step-change in cyber capabilities, and that the change from 90 days to 30 days is a game changer because it allows AI labs to comply without delaying new model releases, synchronizing their efforts under the order with other pre-release activities. After the Times suggested Trump might be micromanaging AI companies, Sacks praised the executive order and specifically addressed concerns that it could become an 'FDA for AI,' noting it expressly forbids creation of a licensing, preclearance, or permitting regime. The Cato Institute, a libertarian institution, offered measured support. Juan Londoño at Cato said the order is imperfect but a step in the right direction, though he expressed concern about vagueness regarding how the government will decide which models qualify for scrutiny. Right-leaning coverage emphasized the order's explicit constraints on government power and its innovation-first framing. Business outlets highlighted that the headline for most companies is what the order does NOT do—mandate government approval before shipping AI—and that the explicit language prohibiting creation of mandatory licensing, pre-clearance, or permitting represented the clearest signal yet that federal posture is innovation-first, with security handled through cooperation rather than restriction. Coverage noted the collaborative rather than restrictive approach, treating AI companies as partners to coordinate with rather than entities to regulate, and that for AI companies preparing trillion-dollar IPOs, the language is reassuring, signaling the federal government won't impose requirements that could undermine valuations or slow product releases. What right-leaning coverage underemphasized were the practical vulnerabilities in a voluntary system—specifically, the lack of clarity about what happens when companies decline to participate, and the significant discretion granted to the NSA director in determining which models qualify as "covered frontier models."
Deep Dive
The story's specific angle centers on how tech industry resistance, particularly from venture capitalist David Sacks, forced a dramatic retreat from what would have been the Trump administration's most significant regulation of AI to date. In early May 2026, the White House was preparing an executive order with a 90-day mandatory review window for frontier AI models before public release—effectively an FDA-style approval process. Trump delayed signing the order in late May over concerns it would interfere with AI innovation, saying he worried it would stifle American companies' lead amid competitive pressure from China, with the earlier version giving the government up to 90 days to review advanced models before release before the timeline was cut to 30 days in the final order. This represents not merely a policy adjustment but a complete inversion of regulatory philosophy—from government-led safety review to industry-led voluntary cooperation. Behind-the-scenes reporting shows Sacks secured a shorter 30-day window and a voluntary framework as some pushed to make it mandatory, also securing agreement that the executive order apply only to advanced models, not incremental versions. The leverage came from the administration's own strategic imperative: Trump delayed the more demanding May version after industry pushback, including from Sacks, because the president didn't want to do anything to get in AI firms' way of leading against China. This was not a policy debate fought primarily on safety merits but on geopolitical grounds. What each perspective gets right and omits reveals important blind spots. The left correctly identifies that a purely voluntary system provides companies with effective veto power over government access and lacks enforcement mechanisms to prevent participation-free launches. The concern about NSA discretion enabling government to weaponize the policy against companies is legally and politically grounded. However, the left underweights the genuine tension between rapid deployment and thorough review in an area where China's investment in AI capabilities is accelerating. The right correctly emphasizes that mandatory pre-approval would likely slow American development and create licensing regimes that benefit incumbents like OpenAI over startups. However, the right minimizes the practical risk that a company facing NSA pressure could experience de facto forced participation despite formal voluntarism, and that 30 days may indeed be insufficient for meaningful cybersecurity threat assessment. Even proponents acknowledge the order does not foreclose potential abuse of the voluntary system, noting that while it makes clear no mandatory process exists, the process could nevertheless be used to place pressure on labs, and the OSTP clarified they are NOT conducting oversight of all new models as that would have chilling effects on free speech and innovation. The unresolved questions going forward are practical and significant: First, will major AI labs actually participate in the 30-day review given the statement is voluntary, and what signals or pressures—formal or informal—might drive participation? Second, will 30 days prove sufficient for the interagency process to identify meaningful cybersecurity vulnerabilities, or will agencies simply rubber-stamp models? Third, what happens when a frontier model launches without government pre-review—will there be consequences, and if so, what? The order implies that labs are expected to go through another pre-deployment phase with trusted partners after the 30-day government evaluation period, but without enforcement teeth, the future of AI regulation becomes a theater production where regulators play audience to tech companies writing their own performance reviews. These operational details, not yet determined, will ultimately determine whether the order achieves its stated goal of balancing security and innovation.