OpenAI has launched a limited preview of its latest artificial intelligence models exclusively to selected partners in the United States, citing a request from the US government as authorities intensify scrutiny of advanced AI technologies.
The rollout marks the latest development in Washington’s growing involvement in the AI sector, particularly over concerns that increasingly powerful models could pose national security risks.
According to OpenAI, access to its new GPT-5.6 series has initially been restricted to a small group of trusted US-based partners whose identities have been shared with government authorities.
The company said overseas employees working for those organisations would also be allowed to use the models.
The move comes weeks after the White House directed AI firm Anthropic to block foreign nationals from accessing its Fable 5 and Mythos 5 models. Anthropic subsequently suspended access, arguing that it could not reliably enforce such restrictions.
Advanced AI systems have recently attracted heightened attention from policymakers due to their reported ability to identify software vulnerabilities that could potentially be exploited by cybercriminals or hostile actors.
Earlier this month, President Donald Trump signed an executive order establishing a voluntary federal review process for assessing national security risks associated with advanced AI models before their public release.While the administration has provided limited details on how the review process will operate, the policy has already sparked debate across the technology industry.
Critics have accused the government of overreach, while supporters argue that safeguards are necessary as AI capabilities continue to accelerate.
OpenAI acknowledged its reservations about the current process but said it decided to cooperate in order to facilitate wider access to the models in the near future.The company said it had briefed government officials on the capabilities of the GPT-5.6 series before launch and agreed to begin with a restricted deployment while broader availability is evaluated.
OpenAI also warned against making government review a permanent requirement for future AI releases, arguing that excessive restrictions could slow innovation and limit access for developers, businesses, researchers and cybersecurity professionals.
The GPT-5.6 family consists of three models designed for different use cases. Sol serves as OpenAI’s flagship model, Terra is aimed at general-purpose tasks, while Luna is positioned as a faster and more affordable option. OpenAI said Terra would cost roughly half as much as its predecessor, GPT-5.5, reflecting growing competition among major AI developers seeking to attract enterprise customers and expand market share.
The latest development highlights the increasingly intense rivalry among leading AI companies, including OpenAI, Anthropic and Google, as they race to develop more capable systems while navigating mounting regulatory and security concerns.The commercial stakes have also risen significantly, with OpenAI and Anthropic reportedly pursuing public listings that could value each company at close to $1 trillion.
As governments and technology firms grapple with the implications of rapidly advancing AI, the balance between innovation, accessibility and national security is likely to remain a defining issue for the industry.
