Silicon Valley killed Trump's AI safety order
Trump was supposed to sign a landmark executive order on AI safety at the Oval Office on Thursday. Tech CEOs had been invited. Some were already on their way to Washington.
Then the phones started ringing.
Between Wednesday evening and Thursday morning, Elon Musk, Mark Zuckerberg, and former AI advisor David Sacks called the president directly. They warned that the planned review system could slow down AI development. By Thursday, the signing was off.
The order would have created a voluntary vetting system. AI companies would submit their frontier models to federal agencies up to 90 days before release. The government would test for dangerous capabilities and find weaknesses before hackers or foreign actors could exploit them. The draft explicitly ruled out mandatory government licensing or pre-approval. Companies could have said no.
That was not enough for Silicon Valley.
The push was partly motivated by new models like Anthropic’s Mythos, which can independently find and exploit security flaws in code. A leaked draft of the shelved order included new AI cybersecurity measures that the White House had prepared before Trump pulled the plug.
The most dramatic reversal came from David Sacks himself. He had been briefed on the draft by science advisor Michael Kratsios, staff secretary Will Scharf, and National Cyber Director Sean Cairncross. He initially said he could live with it. Then, late Wednesday night, he started raising concerns that the voluntary system could become mandatory in practice and be abused by future administrations. By Thursday morning, he called Trump directly without telling anyone, including his own staff, and derailed the whole thing.
A government official told Axios the entire effort was “just something the so-called doomers wanted.” OpenAI lobbyist Chris Lehane broadly supported the order. Other companies pushed to cut the 90-day review window to 14 days. The draft’s plan to give the Treasury Department a lead role also raised eyebrows, since safety reviews are typically handled by CISA and NIST.
Musk has denied responsibility. The White House said Trump had concerns about “aspects” of the order and did not want to do anything that could threaten America’s lead in the AI race with China.
The US now has no federal framework for evaluating AI models before release. Europe has the EU AI Act. China has its own regulatory system. Washington cannot even agree on voluntary testing.
Sources: The Decoder, AFP via Economic Times, NBC News, Malay Mail, US News