OpenAI, a leading artificial intelligence developer, has announced a limited release for its newest suite of AI models, GPT-5.6, following a direct request from the US government. The advanced models, which include the flagship Sol, the balanced Terra, and the cost-effective Luna, are currently accessible only to a small cohort of 'trusted partners' whose involvement has been communicated to the US administration.
This move comes amidst increasing pressure from the US government on AI companies to regulate the deployment of their most powerful systems. The administration previously intervened with Anthropic, another prominent AI firm, by ordering the removal of access for foreign nationals to its Fable 5 model, leading to its complete withdrawal. OpenAI, while complying with the present request, has publicly stated its disagreement with such a restrictive approach.
In a blog post, OpenAI articulated its concerns, stating, “We don’t believe this kind of government access process should become the long-term default.” The company argued that these limitations prevent critical tools from reaching “users, developers, enterprises, cyber defenders, and global partners who need them.” This sentiment underscores a growing debate regarding the appropriate level of government oversight over cutting-edge AI technology.
The current situation appears to stem from a recent US executive order, which encourages certain AI developers to voluntarily submit their advanced models for government review up to 30 days prior to release. Dean Ball, a former White House AI adviser soon to join OpenAI, suggested this has effectively created an involuntary licensing regime for frontier AI, potentially leading to significant launch delays and stifling innovation. Concerns have been raised that this could not only disadvantage the West in the global AI race but also jeopardise substantial investments in AI infrastructure.
OpenAI has described the current limited preview as a “short-term step,” indicating that the GPT-5.6 models are expected to achieve broader availability in the coming weeks. This expanded access is contingent on the company's ongoing collaboration with the US administration to establish a new executive order framework focused on cybersecurity and to develop a consistent process for future model releases.
The GPT-5.6 Sol model is touted as OpenAI's most robust offering to date, featuring enhanced agentic capabilities in areas such as coding, biology, and cybersecurity. It introduces advanced reasoning modes, including “max” and “ultra”, which utilise coordinated subagents to tackle highly complex tasks. OpenAI also highlights Sol's robust security stack, designed to be resilient against adversarial attacks and optimised for defensive cybersecurity applications, prioritising protection against exploits over enabling them.