OpenAI delays GPT-5.6 launch after U.S. government request

OpenAI delayed a full public launch of GPT-5.6 after a request from the U.S. government, limiting early access to a small group of vetted partners while it works with Washington on a broader release framework.

The restricted rollout is temporary, and broader access should follow in the coming weeks. OpenAI had already briefed the government on its plans and the model’s capabilities before launch, and the names of early-access partners were shared with authorities. OpenAI warned that this level of government gating should not become a permanent standard.

President Donald Trump’s June 2 executive order created a voluntary framework for AI developers to provide covered frontier models to the government for up to 30 days before release to trusted partners. The order is meant to pair AI innovation with security and directs federal agencies to work with the private sector on cybersecurity and defenses against external threats.

Officials’ concerns center on national security, especially the risk that advanced models could be adapted for cyberattacks or military misuse before safety teams and regulators have a chance to study them. OpenAI’s GPT-5.6 lineup includes Sol, its most advanced model, along with Terra, a mid-tier version, and Luna, a lower-cost model.

On June 15, Commerce Secretary Howard Lutnick said the government acted against Anthropic’s latest Mythos and Fable models because officials feared they could be used by military intelligence users in China, Russia, or other countries of concern. The U.S. Commerce Department ordered Anthropic to suspend export of the models worldwide and to all foreign nationals, and senior Anthropic technical staff later met with department officials.

Both OpenAI and Anthropic have also confidentially filed for U.S. initial public offerings.