OpenAI announced Friday it will preview its newest GPT-5.6 model series with only a select group of partners before a broader public rollout, following a request from the U.S. government. The company said it shared plans and capabilities for the Sol, Terra and Luna variants with federal officials, who asked that initial access be limited to what OpenAI called "a small group of trusted partners" before wider availability in coming weeks.
The move follows an executive order President Trump signed earlier this month laying out a voluntary testing process allowing AI labs to provide models to the government up to 30 days ahead of release for risk assessment. The administration has emphasized that participation is not mandatory, but recent actions suggest otherwise. Earlier this month, the Trump administration sent OpenAI competitor Anthropic a directive to pull its newest Fable and Mythos model over security concerns within hours of receiving a federal export control order.
What the Right Is Saying
Administration officials and their allies in Congress have defended the approach as a reasonable precaution to address legitimate cybersecurity threats. National security hawks have pointed to foreign adversaries' interest in accessing cutting-edge AI capabilities, arguing that some form of pre-release review serves American interests. Supporters note the executive order explicitly frames testing as voluntary and argue it establishes a workable framework for government-industry coordination without imposing binding restrictions.
Conservative commentators focused on technology policy praised what they characterized as an attempt to balance innovation with security. Dean Ball, a senior fellow at the Foundation for American Innovation who co-authored Trump's AI Action Plan released last year, initially described the executive order as "really establishing a de facto involuntary licensing/preapproval regime for frontier models." He subsequently wrote that within weeks, federal AI policy had shifted from what he called an implausibly libertarian approach to one increasingly draconian and opaque.
What the Left Is Saying
Democratic policy advocates and AI researchers have raised concerns about what they describe as an opaque and unpredictable approach to AI governance. Critics argue that ad hoc directives to individual companies create regulatory uncertainty that could stifle innovation at a critical juncture for American competitiveness in artificial intelligence. Several technologists noted that the lack of clear, codified standards makes it difficult for labs to plan development cycles or understand compliance requirements.
The backlash from the Anthropic directive sparked warnings from AI policy advocates who said the episode signaled the White House was taking an overly discretionary approach to model releases. Some progressive tech-policy groups have argued that meaningful AI safety oversight requires transparent rulemaking with public comment periods, not case-by-case government intervention. The absence of clear guidelines, they contend, disproportionately advantages large established firms with regulatory compliance teams over smaller research organizations.
What the Numbers Show
The executive order signed earlier this month requires companies participating in the voluntary testing process to provide government access to new models up to 30 days before public release. The Trump administration has not specified how many companies have participated in the program since its inception. OpenAI's announcement marks the first major instance of a leading AI lab publicly acknowledging it adjusted its rollout timeline at federal request, though the company characterized this as a "short-term step." Anthropic complied within hours when directed to disable access to its Fable and Mythos models for foreign nationals.
The Bottom Line
The staggered release reflects ongoing tension between the Trump administration's stated goal of maintaining American AI leadership and its actions to restrict model availability during a testing period. OpenAI explicitly pushed back against what it called government access becoming a "long-term default," writing that such processes "kee[p] the best tools from users, developers, enterprises, cyber defenders, and global partners who need them." The company said it views this as "the strongest path to broader availability in the coming weeks" while working with the administration on developing a more permanent framework. Industry observers will be watching whether other major AI labs face similar requests ahead of planned releases.