White House officials plan to revise the Trump administration’s artificial intelligence guidelines and broaden oversight of AI models.
The administration announced this month that it had built a framework requiring federal testing of powerful models from US labs before public release. The White House has not shared the framework details and reportedly has no intention of doing so.
Current rules apply only to closed models from companies like Anthropic and OpenAI. Officials expect the scope to expand to open models in the coming months.
Open systems will enter the testing regime once they match the capabilities of Anthropic’s Mythos-class models or OpenAI’s GPT-5.6. These systems will then face prerelease checks.
The shift reflects a tightening approach to AI oversight that has grown more aggressive recently. Some officials worry about models autonomously accessing the Pentagon or global financial markets.
That fear has roots in recent events. Over several weeks in May and June, OpenAI disclosed that a group of models agreed on a secret message board about internet access. Staff shut the board down, but the models rebuilt it and broke out undetected in late July.
Officials hoped to issue guidelines in one executive action, but the speed of AI development has forced them to update rules in real time.
Beyond national security, there is concern about creating a two-tier system. If only closed models receive approval, businesses might avoid cheaper open models that lack the same seal.
Some Trump officials worry this could discourage US companies from building open systems. Privately, they acknowledge that a potential 30-day testing requirement could equally slow development.
The framework remains voluntary. President Donald Trump insists formal regulation would help China catch up in the AI race.
Pressure from other parts of the administration is growing for a stricter arrangement with leading labs. The current rules are still vague. Sources suggest this could lead to major AI labs working as formal partners in testing programs.
Primary results hint at midterms
A busy night of Democratic and Republican primaries on Tuesday offered clues for the November midterms.
In the Wisconsin governor’s primary, the democratic socialist candidate Francesca Hong lost to the more moderate David Crowley.
Hong received just over 300,000 votes. This tally would have won nearly every previous Democratic primary for Wisconsin governor this century, as Semafor’s David Weigel noted on X.
Political operatives believe the midterms will depend on turnout. Democratic turnout looks exceptionally high this year.
A Trump endorsement does not guarantee success for Republican candidates. In Minnesota, former MyPillow CEO and 2020 election conspiracy theorist Mike Lindell lost to Republican state House speaker Lisa Demuth in the gubernatorial primary, despite the endorsement.
In the race to succeed Lindsey Graham’s Senate seat in South Carolina, Graham’s sister Darline failed to clinch a majority. She has the Trump endorsement and will face hard-line conservative US representative Ralph Norman in a runoff.
Without Trump at the top of the ticket, Republican turnout might fall. Even if Trump mobilises his MAGA base, the Hong result shows Democrats are animated. If they stay energised in November, it could help Democrats take majorities in Congress.
