Trump yanked AI order after David Sacks raised industry concerns

Thursday’s abrupt postponement of President Donald Trump’s much-awaited executive order on artificial intelligence came after former AI czar David Sacks voiced industry concerns about the measure to Trump, according to a senior White House official and two people familiar with the matter.

Sacks’ 11th hour intervention — and his arguments that the order could prove too onerous for the rapidly evolving AI industry — came even though he had been briefed about the directive in recent days, one of the people told POLITICO. The people were granted anonymity to discuss internal conversations.

The executive order, which the White House planned to release Thursday afternoon, would have set in motion a voluntary oversight system in which developers of advanced AI models could submit their products to a review by federal agencies before releasing them, POLITICO previously reported. It was meant to address concerns that advanced AI products from companies like Anthropic could unleash devastating cyberattacks and other havoc if they fell into the wrong hands.

But during a conversation with Trump, Sacks told the president that companies were already cooperating, and that having the federal government review models before their public release would slow down innovation and harm the U.S. in its AI race with China, the senior White House official and one of the other people said.

Sacks did not respond to a request for comment.

The Silicon Valley venture capitalist was not the sole obstacle, one of the people said, citing opposition from some other industry leaders.

The proposal had gotten a mixed reception in the tech industry: OpenAI, developer of one of the most advanced AI models, has been supportive of the contours of the order.

Chris Lehane, OpenAI’s top lobbyist, told reporters last week that the company believed in the importance of collaborating with the government on AI safety. Lehane said the company was focused on “the ability to innovate, but doing it in concert with the government in a way that actually really prioritizes that safe deployment.”

Trump offered few details Thursday about why he delayed the signing of the order. The people familiar with the issue said his staff had not only briefed tech executives on the order but had invited several leading Silicon Valley leaders to the ceremony.

“I didn’t like certain aspects of it,” Trump told reporters Thursday morning. “I think it gets in the way of — we’re leading China. We’re leading everybody, and I don’t want to do anything that’s going to get in the way of that.”

But the senior White House official and the two people said the rollout unraveled after Sacks raised objections directly with Trump in the days before its planned release, blindsiding some White House staff involved in the process.

According to the White House official, Sacks had participated in a review of the EO this week, and White House officials believed he was generally happy with it and would support it.

But Wednesday night, he began to raise concerns, including fears that the voluntary nature of the agreement may one day become mandatory, the senior White House official said.

“Then, he called POTUS this morning unbeknownst to anybody, his own staff included, and derailed it,” the White House official said.

The reversal also came after industry officials raised concerns about a proposed voluntary review process for cutting-edge “frontier” AI models, according to four people familiar with the matter, who were granted anonymity to discuss private discussions. Industry officials had pushed for the White House to shorten the time frame to share new models with the White House from 90 days to 14 days, said three of the people.

Industry officials also pushed for the proposed order to give responsibility to the intelligence community to lead the effort to review new models, according to one of the people. A readout of the order shared with industry officials on Tuesday would give the National Security Agency final say on which systems are considered “covered frontier models,” with other agencies helping set up a classified benchmarking process within 60 days.

A draft of the order shared with POLITICO stressed the voluntary nature of the review: “Nothing in this section shall be construed to authorize the creation of a mandatory governmental licensing, preclearance, or permitting requirement for the development, publication, release, or distribution of new AI models, including frontier models.”

Four people familiar with the planning also said the delay was due in part to the fact that several leading tech CEOs who were invited on short notice could not attend. OpenAI, Anthropic and Meta were all expected to send lower executives in place of their CEOs, people familiar with planning said.