OpenAI Launches Limited GPT-5.6 Preview Amid U.S. Security Review

OpenAI has begun a limited preview of its next-generation model, GPT-5.6. The rollout targets a small group of trusted researchers, enterprise partners, and safety evaluators through the company API. The move occurs while OpenAI continues to participate in a U.S. government security review examining advanced AI systems for national security risks.

The preview gives participants early access to test the model in controlled environments. OpenAI described the step as a deliberate choice to gather detailed feedback and refine safety measures before any wider release. Company materials state that GPT-5.6 delivers stronger performance on long-horizon reasoning tasks, more reliable code generation, and improved handling of complex multimodal inputs compared with earlier versions.

Officials involved in the security review have focused on how frontier models handle sensitive domains such as biological research assistance, cybersecurity applications, and autonomous decision-making. The review process requires OpenAI to share technical documentation and demonstrate controls against misuse or unauthorized access by foreign entities. OpenAI has confirmed full cooperation and noted that additional safeguards were added specifically for the preview phase.

In prepared remarks, OpenAI CEO Sam Altman said the company remains committed to advancing AI capabilities in a manner that supports U.S. technological leadership while addressing legitimate security concerns. He added that phased releases help identify issues early and allow iterative improvements based on real-world testing data.

Industry observers noted that the limited preview balances OpenAI push for rapid progress with growing regulatory scrutiny. Several analysts pointed out that similar reviews now affect multiple leading AI developers as governments seek to manage risks associated with increasingly powerful systems. Some researchers in the preview program have already reported gains in areas such as scientific hypothesis generation and multi-step planning, though they emphasized the need for continued evaluation.

Access remains restricted. OpenAI has not announced pricing or a timeline for broader availability. The company stated that expansion will depend on internal safety assessments and the outcome of the ongoing government review. Further updates are expected once additional testing data is analyzed.

The announcement highlights the current tension in AI development between innovation speed and oversight requirements. OpenAI positioned the GPT-5.6 preview as evidence of its continued focus on responsible deployment amid these evolving conditions.