In a characteristic eleventh-hour reversal, President Donald Trump scrapped a sweeping executive order on AI safety following a high-stakes lobbying blitz by Elon Musk, Mark Zuckerberg, and David Sacks. The White House had already distributed invitations for a formal Oval Office signing ceremony scheduled for Thursday, only to pull the plug after a series of frantic phone calls from Silicon Valley’s new political vanguard. According to reports from the Washington Post and Politico, the tech trio warned Trump that the proposed safeguards were less about safety and more about surrendering America’s lead to China in a fit of regulatory self-sabotage.
The friction point wasn’t just the paperwork; it was the precedent. The draft proposed a vetting mechanism where companies would 'voluntarily' hand over frontier models to federal agencies for a 90-day testing period before public release. While the administration framed this as a light-touch approach, figures like David Sacks saw a Trojan horse. Sacks argued that such frameworks inevitably harden into mandatory licensing regimes, ripe for weaponization by future administrations. The sudden pivot has left OpenAI’s lobbyist Chris Lehane and other safety advocates in the lurch, signaling that the era of the 'AI doomer' influence in Washington is effectively over.
The administration’s internal logic, as cited by Axios, now reflects a ruthless prioritization of market speed over independent risk auditing. Officials have dismissed the previous safety-first rhetoric as an unnecessary concession to existential-risk theorists who threaten to stall progress. By choosing the accelerationist path, Trump has signaled that the U.S. will no longer entertain European-style precautionary principles. In this new paradigm, the only unacceptable risk is being second to Beijing, and if that requires bypassing audits for models like Anthropic’s vulnerability-hunting Mythos, the administration seems perfectly willing to pay that price.