The reported rollout marks a departure from OpenAI’s previous flagship model launches, which typically expanded from internal testing to enterprise previews and broader availability over a relatively short period. Under the new approach, GPT-5.6 access will be granted to selected enterprise customers before a wider release.
Neither OpenAI nor the administration has publicly identified the security concerns behind the request. The Information reported that the federal government asked the company to stagger the rollout, while additional reporting said customer approvals during the preview period will be handled on a case-by-case basis.
The reported changes come weeks after Anthropic said it was forced to suspend access to its Fable 5 and Mythos 5 models following a U.S. government export-control directive. Anthropic said the order applied to foreign nationals, including non-U.S. citizen employees working inside the United States, and required the company to disable both models for customers to comply with the directive.
In its statement, Anthropic said the government did not provide specific details about the national security concern and that it understood the issue involved a possible jailbreak of Fable 5.
The company also defended its testing process, saying it had worked with government agencies and outside organizations before launch. Anthropic argued that similar capabilities existed in other publicly available models, including OpenAI’s GPT-5.5, and said applying the same standard across the industry would “essentially halt all new model deployments for all frontier model providers.”
The Information also reported that OpenAI, Meta and Reflection AI participated in a June 9 feedback session hosted by the Office of the National Cyber Director, while Anthropic was not invited.
The reported GPT-5.6 rollout follows the Trump administration’s June 2 framework for voluntary pre-release cooperation with developers of frontier AI models. According to the reports, the framework allows companies developing covered models to provide the government with pre-release access without establishing a formal licensing system.
Altman has previously supported greater federal oversight of advanced AI systems. According to the Center for Strategic and International Studies, he told Congress in 2023 that the United States should create a federal agency with authority to issue, deny and revoke licenses for frontier AI development.
For OpenAI, the limited preview allows GPT-5.6 to begin reaching customers while complying with the reported government request. For enterprise customers awaiting the model, broader availability will depend on how long the preview period lasts and how customer approvals are handled, neither of which has been publicly detailed.
This analysis is based on reporting from runtimewire.
Image courtesy of webiano.digital.
This article was generated with AI assistance and reviewed for accuracy and quality.
