The Trump administration’s AI policy was rewritten in a late-night phone call. Hours before a scheduled signing ceremony, the president pulled the original draft after David Sacks, the White House AI czar, warned it would cede the technological race to China.
The final order, signed quietly on a Tuesday, mandates a voluntary 30-day safety review instead of the proposed 90-day window. It includes an explicit disclaimer prohibiting any mandatory government licensing or pre-clearance regime. On The AI Daily Brief, host Nathaniel Whittemore described the process as a chaotic negotiation, with the administration treating regulation like "eating its vegetables" - creating a paper trail for oversight while protecting tech interests.
"The final version cuts that window to 30 days and explicitly forbids the government from creating a mandatory licensing or permitting regime."
- Nathaniel Whittemore, The AI Daily Brief
The policy shift was triggered by Anthropic’s April announcement of its Mythos model, which the company deemed too dangerous for public release due to its skill in finding software vulnerabilities. Microsoft and JPMorgan Chase CEO Jamie Dimon then warned the White House, spurring Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent to push for guardrails against potential cyberattacks.
Internally, the administration fractured. Libertarian advisors like Sacks clashed with pragmatists Wiles and Bessent, who viewed total lack of oversight as a political liability. After tech executives including Mark Zuckerberg and Marc Andreessen called Trump directly, he briefly canceled the signing. The compromise, granting the National Security Agency primary testing responsibility but no power to block a release, is a tactical retreat that leaves the door to future regulation ajar.
"The door to federal regulation is no longer locked."
- Trip Mickle, The Daily
While Washington negotiates, the enterprise landscape is shifting under a new constraint: cost. Microsoft’s new MAI models prioritize being ten times cheaper over winning raw performance benchmarks, targeting what CEO Satya Nadella calls a "token shortage." Companies like Uber are implementing hard caps, limiting employee token spending to $1,500 per month to manage the spiraling expense of agentic workloads.
Concurrently, OpenAI is redesigning knowledge work itself. Codex now has 5 million weekly active users, with non-technical employees adopting it three times faster than developers. Half of its users now run multiple tasks in parallel, orchestrating workstreams rather than executing them sequentially. New features like "Sites" turn reports into disposable web apps, a bet that interactive tools will become the standard for corporate coordination.
The executive order formalizes a voluntary handshake, but the underlying economic and technological forces are moving faster than any policy process.
