OpenAI delays GPT-5.6 after government request

According to The Information, the Trump administration asked OpenAI to stagger the rollout of its next major model because of security concerns; Reuters and The Verge report the same development. Reuters reports that OpenAI CEO Sam Altman told employees the company would release GPT-5.6 in a limited preview to select enterprise partners, and that the government would "approve access customer by customer during this preview period," citing The Information. Reuters further reports the request came from the Office of the National Cyber Director and the Office of Science and Technology Policy, according to The Information. The Verge contrasts the arrangement with an earlier directive that restricted non-US access to Anthropic's Mythos 5 and Fable 5.

Industry-pattern observations: Major model releases increasingly trigger cross-functional reviews involving security, export-control, and national-security offices. Companies rolling out frontier models face a mix of technical gating (limited previews, enterprise-only access) and policy gating (government-led customer approvals) as near-term controls. For practitioners, these combined gates typically slow broad access, concentrate usage data in early partner cohorts, and raise operational overhead for access management and compliance.

Government requests to stagger model rollouts represent a significant intersection of national security policy and AI deployment practice. Reporting by Reuters and The Verge places this event alongside prior regulatory moves affecting Anthropic, suggesting a broader executive-branch focus on controlling external access to high-capability models. For practitioners building on or deploying large models, this trend changes the timeline and scope for early integrations, third-party testing, and cross-border collaborations.

Observers should track three signals. First, reporting from The Information and Reuters indicates whether the government publishes formal guidance or directives following this request. Second, monitor which enterprise partners receive access during the GPT-5.6 preview and whether access conditions (geography, citizenship, sector restrictions) are publicly disclosed. Third, watch for changes to contractual or API-level controls that vendors roll out to support customer-by-customer approvals, since those controls affect deployment automation and compliance tooling.

For practitioners evaluating model integrations, the immediate effect is likely narrower early access and potentially stricter eligibility checks for test integrations. Industry-pattern observations: When access is limited and approved case-by-case, organizations that obtain preview access often need to support extra compliance and security reporting, and vendor rate-limiting or contractual constraints can increase implementation friction. Public reporting does not document OpenAI's internal compliance processes; The Information and Reuters are the sources for the announced preview arrangement.

All reporting in this item attributes the request and the planned limited preview to The Information, as summarized by Reuters and The Verge. OpenAI has not been quoted here beyond staff statements attributed in those reports.