The Trump administration has reportedly asked OpenAI to roll out its latest model, GPT-5.6, in stages, citing cybersecurity concerns. For now the model will be limited to a preview for a small number of partners.
June 25, 2026 · Frontier AI Policy
White House Asks OpenAI to Stagger GPT-5.6 Release on Security Grounds
Instead of a single broad launch, the government is reportedly requiring a "customer-by-customer" approval during preview — turning frontier models into tightly controlled geopolitical assets.
~90%
Prediction-market odds of a release by month's end
$1.1M+
Wagered on a June 22–28 launch window
~6 wks
OpenAI's usual flagship update cadence
Rumored upgrade · unconfirmed
Context Window: 1M → 1.5M tokens
A 50% larger context plus a refreshed ~May 2026 training cutoff and a redesigned alignment stack.
The New "Customer-by-Customer" Rollout
Preview build
Codename "kindle-alpha"
→
Gov't review
Approval per customer
→
Staged access
Eventual broad release
Normally API access follows a ChatGPT launch within 24–48 hours. This time the timeline is uncertain, expanding only as approvals clear.
Agreed to voluntary review
Google, xAI and Microsoft are said to have signed on. A precedent: the government ordered Anthropic to halt foreign-national access to Fable 5 and Mythos 5 over jailbreak risk.
Resistance & concerns
Only Meta is reportedly resisting. Critics warn that blocking all jailbreaks may not be technically possible, and that normalized controls could delay any broad release and dent competitiveness.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…