Breaking

OpenAI to Limit GPT-5.6 to About 20 Partners as US Administration Seeks Pre-Review

June 25, 2026 at 22:42 EDT

Foundation Models
Policy & Regulation
AI Agents

OpenAI plans to give its next flagship model, "GPT-5.6," first to about 20 to 24 government-approved partners before a broader rollout, according to reports. The move follows a request from the Trump administration for prior review and a staggered release, citing security concerns.

Frontier AI · Government Oversight

GPT-5.6 Goes to ~24 Partners First — Under Government Vetting

For the first time, the US administration intervened in advance over a frontier model, asking OpenAI to stagger the release and vet every user and partner — granted "customer by customer," each requiring government approval.

~24

Partners with early enterprise-preview access

~900M

ChatGPT general users with NO access yet

1st

Pre-release government intervention on a frontier model

Access Gap

Two dozen partners vs. the wider audience locked out

~24

partners

~900M

general users (no access)

Delivered via API and partner-only previews — not to ChatGPT's general base. A broader release is expected within weeks, paced by approvals.

Model Timeline

GPT-5

Aug 2025

General release

GPT-5.5

Apr 2026

General release · 400K context

GPT-5.6

TBD

~24 partners, then staggered

Benchmarks, pricing, and context window remain undisclosed; no formal announcement made.

Optimistic view

Expected gains in reasoning and agentic capabilities
Aim to harness AI's power while limiting national-security risk

Concerns

Frustration over release delays and restrictions
Stronger vetting could erode US AI leadership
Room for Chinese open-weight rivals — Kimi, GLM-5.2, Qwen

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.