OpenAI Releases New AI Models Under Strict Government Supervision

OpenAI has officially released its new GPT-5.6 model suite, but the launch comes with significant restrictions. Following a request from the US government, the company is using a "staggered" release plan. This means the AI models are not being made available to everyone at once. Instead, they are being shared with a small group of trusted partners first. According to The Guardian, this move is similar to a recent situation involving a rival company, Anthropic, which had to stop public access to its own technology due to government security concerns.

The new GPT-5.6 suite includes three different versions: Sol, Terra, and Luna. The Verge reports that Sol is the most powerful model in the group, while Terra is designed for high-volume work, and Luna is the most affordable option. OpenAI claims that Sol is its strongest model yet, specifically mentioning its skills in coding, biology, and cybersecurity. However, the company also emphasized that the model is designed to help people find and fix security problems rather than perform dangerous cyber attacks.

This limited release is happening because of an executive order issued by President Donald Trump. The order requires companies to follow a specific framework for checking and deploying new, powerful AI models. OpenAI stated that it is working closely with the White House to develop these safety rules. However, the company also made it clear that it does not want this type of government oversight to become a permanent requirement. In a blog post, OpenAI noted that this process prevents the best tools from reaching users, developers, and global partners who need them. Despite these concerns, the company is following the rules for now, hoping that this short-term step will lead to broader availability in the coming weeks.

During this preview period, the US government is approving customers on a case-by-case basis. OpenAI’s chief executive, Sam Altman, told staff that the government would be monitoring the process closely. If everything goes well, the models will be released to the general public in about two weeks. For now, access is limited to US-based entities, though employees of those companies who are based in countries like the UK and Australia may also be able to use the technology.

While OpenAI manages these regulatory hurdles, it is also facing new developments in a major legal case. The New York Times is currently updating its lawsuit against OpenAI and its partner, Microsoft. In a recent court filing, the newspaper proposed changes to its complaint, arguing that Microsoft actively encouraged OpenAI to use copyrighted works without permission. The New York Times claims that Microsoft built a powerful supercomputer specifically to help OpenAI train its models on this protected content.

According to Ars Technica, the legal move by The New York Times follows a change in how courts look at "contributory infringement." This is a legal term for when a party is accused of helping others break the law. The newspaper wants to align its case with new legal standards, which require plaintiffs to prove that companies intentionally acted to cause illegal conduct. A spokesperson for The New York Times stated that their core argument remains the same: that Microsoft and OpenAI used millions of copyrighted works to build products that compete with the newspaper.

In response, a Microsoft spokesperson described the updated lawsuit as a "last-ditch effort" to save a case that has faced unfavorable rulings in other recent legal battles. Microsoft maintains that the claims are without merit. As the situation develops, OpenAI finds itself in a difficult position. It must satisfy the strict safety demands of the US government to keep its new models moving toward a full release, while simultaneously defending its business practices in court. The coming weeks will be critical for the company as it attempts to balance these government requirements with its goal of providing broad access to its latest AI tools.