Sam Altman told OpenAI staff this week that the White House is controlling how the company releases its new AI model.
According to TechCrunch reporting, the OpenAI CEO revealed that GPT 5.6 will not receive a public rollout like previous releases. Instead, access goes to select partners only, with the Trump administration approving each customer individually.
The Office of the National Cyber Director and the Office of Science and Technology Policy both requested the limited release, citing concerns about frontier AI capabilities reaching adversaries.
Altman told staff the preview period could last "a couple of weeks," with hopes for a broader public release afterward if things proceed as planned. The White House effectively controls OpenAI's release timeline.
This marks a reversal for the Trump administration, which initially portrayed itself as hands-off on AI regulation. This month, Trump signed an executive order requiring certain AI companies to submit new models to the government for testing before public release.
OpenAI is not alone. Rival Anthropic already restricted its Claude Mythos model through Project Glasswing, releasing it only to a select group. That decision generated debate about whether the company was exercising genuine caution or pursuing marketing advantage.
The underlying concern is substantial: frontier AI models can identify and exploit software vulnerabilities faster than human analysts. In hostile hands, that creates serious risks. Cybercriminals already use AI to write malware and execute autonomous ransomware attacks.
OpenAI now operates under government oversight. The AI sector has gained a regulator.
