The eleventh-hour decision by President Donald Trump to indefinitely postpone an executive order concerning artificial intelligence brought an unexpected turn of events yesterday. Executives from major technology companies, some already en route to Washington for the anticipated signing ceremony, found themselves in limbo after Trump cited concerns about hindering U.S. leadership in AI development. This sudden reversal followed a concerted lobbying effort from prominent tech billionaires, including Elon Musk, Mark Zuckerberg, and venture capitalist David Sacks, who reportedly contacted Trump directly to express their apprehensions.
The proposed executive order aimed to establish a voluntary system allowing AI companies to submit their most advanced models, such as Anthropic’s powerful Mythos, to national security agencies for testing and vetting up to 90 days before public release. This framework had been under discussion for weeks among various government departments, with companies like Anthropic and OpenAI reportedly indicating their support for the voluntary vetting process. Anthropic, in particular, has voluntarily restricted the wider release of its Mythos model, acknowledging its unprecedented cyber capabilities could pose significant risks if misused.
David Sacks, who previously served as Trump’s AI and crypto czar, emerged as a central figure in derailing the order. A vocal AI “accelerationist,” Sacks champions the view that federal regulation stifles innovation, disadvantages American tech companies, and risks ceding ground to rivals like China in the global AI race. Despite no longer holding an official position, Sacks continues to influence the administration’s AI policy, even attending briefings on the executive order earlier this week. Reports indicate that while initially not opposed to the voluntary testing, Sacks later conveyed to Trump fears that it could evolve into a de facto licensing regime, slow down AI model releases, and potentially become mandatory under future administrations.
This intervention highlights a growing ideological chasm within Trump’s support base regarding AI policy. While a faction of Silicon Valley billionaires, characterized by a “move fast and break things” ethos, has found common ground with Trump’s anti-regulatory stance, a significant portion of the MAGA faithful appears to favor government oversight. Polls suggest a majority of Americans, including many Republicans, express anxiety about AI’s potential societal impacts, ranging from job displacement to concerns over education and mental health. This sentiment extends to religious conservatives who view advanced AI with suspicion, sometimes equating it with idolatry.
Indeed, prior to the executive order’s postponement, many in Washington believed its passage was all but certain, with proponents of regulation gaining momentum. Over 60 MAGA loyalists, including former Trump advisor Steve Bannon, signed an open letter urging Trump to mandate testing and approval for powerful AI models before their release. A recent poll from the Future of Life Institute, an AI safety advocacy group, indicated that 79% of Republican voters support government testing of AI models for safety, with 87% favoring governmental authority to block models posing national security threats. This suggests a disconnect between the views of some influential tech figures and the broader conservative electorate on this issue.
The current situation leaves U.S. AI policy in an ambiguous state. While the broader executive order is on hold, an ad hoc licensing process continues for Anthropic’s Mythos model. Through “Project Glasswing,” Anthropic has shared the model with the U.S. government and a select group of technology and financial institutions critical to internet infrastructure. However, the White House has reportedly limited the expansion of this access due to national security concerns. Meanwhile, OpenAI’s GPT-5.5, also highly capable, has seen a less restrictive “trusted access” program without the same level of governmental scrutiny applied to Mythos. This patchwork approach underscores the ongoing challenges and internal disagreements within the administration on how to balance innovation with the profound risks presented by advanced artificial intelligence.
