A sudden warning sign inside the Trump administration triggered days of round-the-clock alarm among senior officials, according to people familiar with internal discussions.
The trigger, officials say, was the preview and staged rollout of Anthropic's powerful Mythos model, which set off fears inside the administration that the technology could be weaponized to launch damaging cyberattacks or otherwise threaten national security, reported the Wall Street Journal.
One senior administration official described the incident as a "red flashing light," prompting officials to convene urgent calls that stretched around the clock for days.
In May, White House AI adviser David Sacks persuaded President Trump to scrap a planned executive order that would have subjected AI models to extended government review — a last-minute reversal that blindsided and angered officials including chief of staff Susie Wiles and Treasury Secretary Scott Bessent, according to the report.
Bessent had worried that AI-enabled cyberattacks could destabilize the banking system. Wiles and Bessent later pushed Trump to sign a scaled-back version of the order instead.
Since then, Wiles, Bessent and other officials have met almost daily to weigh risks and possible safeguards, people familiar with the matter said, even as Trump publicly waves off calls for tighter regulation.
The concern deepened in July, when OpenAI's systems were involved in an incident affecting the company Hugging Face using hundreds of coordinated AI agents, reportedly intensifying fears among officials about models operating outside human control.
White House teams, including Vice President JD Vance, have examined how hostile actors or foreign governments might exploit the technology to target infrastructure such as power grids and water systems, with input at times from the National Security Agency and CIA.
Discussions have also touched on the potential for the technology to aid in the creation of biological weapons, drawing in officials from the Department of Health and Human Services, the Journal reported.
Despite the internal alarm, Trump has remained committed to accelerating AI development, arguing the U.S. must outpace China, according to the report.
He has reportedly dismissed suggestions of a coordinated slowdown and continues to back the industry's rapid buildout of data centers, even as polling shows most Americans oppose such projects in their own communities.
Trump's inner circle is split on the issue, with some aides are pushing for firmer guardrails, while allies including Sacks, Meta's Mark Zuckerberg and Nvidia's Jensen Huang continue to press the president toward a hands-off approach — one that, so far, appears to be prevailing.
