OpenAI announced on Friday that it will postpone the full public release of its GPT-5.6 model following a request from the US government. The initial access to this AI model will be limited to a selected group of partners who have undergone scrutiny and whose details have been shared with the authorities.
This decision underscores growing concerns in Washington regarding the risks posed by powerful artificial intelligence systems on national security. Policymakers are urging companies to implement controls to manage these risks.
US officials aim to secure early access to advanced models to identify threats ranging from cyberattacks to military misuse before these tools are widely deployed.
OpenAI mentioned in a blog post that the limited release is a temporary measure while they work with Washington to establish a broader framework for future releases. The developer of ChatGPT stated that they had presented their plans and model capabilities to the government prior to the release.
OpenAI noted, "We are taking this short-term step as we believe it is the most effective way to enable broader model availability in the coming weeks, while we collaborate with the US administration to develop an executive framework for cybersecurity and a repeatable process for future model releases."
The company added that they will continue rigorous testing and close coordination with partners in preparation for a wider release, but warned that this level of government access and oversight should not become a permanent norm. The company did not disclose the names of its partners.