OpenAI will let the United States government test its most powerful AI models up to 30 days before public release, complying with a voluntary order President Trump signed this week.
Key Points:
- OpenAI agreed to share its top AI models with U.S. regulators before they reach the public.
- The order asks companies to hand over covered frontier models 30 days ahead of release.
- Critics call the voluntary framework too weak to police dangerous systems.
OpenAI Backs Federal Review
George Osborne, OpenAI's head of countries, told CNBC the company would sign up to the framework rather than wait to be asked by regulators. He said it is "quite right that democratic governments have a big role to play" in shaping how the technology is deployed.
Trump signed the order Tuesday, capping months of debate over how to police the most capable systems without choking innovation. It asks firms to submit covered frontier models for a federal benchmarking process run by security agencies.
Officials would build a classified test of cyber capabilities and flag any model that needs tighter controls before it can be sold widely.
That request stops well short of any mandatory licensing or preclearance rule, leaving participation entirely optional.
Also Read: Bitcoin Suffers Its Worst Week As Saylor Blames The AI Trade
Critics Question Trump Order
Not everyone welcomed the move. Rep. Don Beyer, a Virginia Democrat who co-leads a congressional AI caucus, said the order still lets companies ship powerful models to the public without meaningful evaluation.
He argued the framework leaves regulators short of the tools they would need to catch a genuinely dangerous system. Because cooperation stays voluntary, its real impact hinges on how many leading developers choose to opt in. OpenAI had floated a similar review plan of its own earlier this year, urging evaluation without handing officials the power to block a launch.
Trump Order's Rocky Path
An earlier version of the order gave the government 90 days to study new models, a timeline later trimmed to 30. Tech figures including David Sacks and Elon Musk reportedly warned that a longer review could slow American development. A planned signing ceremony with industry executives was scrapped at the last minute before the softer order emerged.
Trump had said he worried the original plan would dent the country's lead over China in artificial intelligence. The final order grew out of alarm over Anthropic's Mythos model, which exposed cyber weaknesses at unusual speed and rattled both governments and Wall Street.
Read Next: Who Really Influences Dogecoin? 10 X Accounts Worth Following