The ChatGPT-maker is launching a limited preview of the models Sol, Terra, and Luna, with Sol described as the new flagship model. The decision marks the first time OpenAI has staggered a flagship release at the request of the U.S. government, according to reports. [1][2]
Previously, the company released new generations of ChatGPT according to its own schedule and directly to the public. The restriction comes amid heightened government scrutiny of advanced artificial intelligence systems and national security risks.
Government Scrutiny of AI Models
The action follows a Trump executive order in June that established a framework for vetting advanced artificial intelligence (AI) systems for national security risks. The initiative, called the Genesis Mission, aims to fast-track U.S. AI dominance by comparing its urgency to the Manhattan Project, according to an article by Kevin Hughes on NaturalNews.com. [3]
Earlier this month, Anthropic disabled access to its Fable 5 and Mythos 5 models three days after launch, after receiving a U.S. government directive to take them offline, according to reports. The order came after Amazon’s own researchers found a jailbreak. [4]
White House officials said they are collaborating with frontier AI labs on addressing the challenges of scaling that technology, though no specific details have been released.
Criticism of Government Intervention
Cybersecurity experts, including Stanford University’s Alex Stamos, said the government’s actions lack a factual basis and may hurt U.S. competitiveness against China. Stamos called the intervention "the dumbest thing they could do" if the goal is to beat China, according to a report. [2]
Investor David Sacks said Anthropic’s CEO raised concerns that led to heightened scrutiny, describing Mythos as having advanced cyber capabilities. Some analysts argue that restricting access to frontier models slows innovation and cedes ground to Chinese AI developers, who have already replicated advanced U.S. models. [11]
In December 2024, researchers from Fudan University and the Shanghai AI Laboratory successfully replicated OpenAI’s o1 reasoning model, according to another article from NaturalNews.com. [5]
OpenAI's Safeguards and Negotiations
OpenAI said the Sol model is better at finding vulnerabilities than carrying out cyberattacks and does not cross its risk threshold, according to the company's announcement. [1] The company acknowledged that unforeseen risks remain and said it is implementing a phased release with stronger safeguards.
CEO Sam Altman spoke with Commerce Secretary Howard Lutnick about the release, according to a report on OpenAI’s agreement with the Department of War (DOW). [6] The talks are part of ongoing discussions between AI executives and Trump administration officials over national security concerns.
OpenAI has also agreed to a $200 million contract with the Pentagon to provide AI tools for battlefield strategy and cybersecurity, according to an article by Belle Carter on NaturalNews.com. [7]
Implications for U.S. AI Competitiveness
The restrictions have drawn criticism from those who argue they slow U.S. innovation and cede ground to Chinese AI developers. Some observers note that China is emerging as a leader in AI research and that open-source and cost-effective models from Chinese companies rival U.S. giants, as reported in an article by Willow Tohi. [8]
OpenAI said it views the testing period as temporary, but no end date has been set, according to the company. [1]
The situation reflects a tension between national security concerns and the push for rapid AI advancement. Some experts have warned that unchecked AI development could lead to catastrophic risks, as noted by former OpenAI researcher Steven Adler. [9]
The role of government oversight in balancing security and innovation remains unsettled.
References
- OpenAI Releases New Flagship Model to Limited Users Following Government Concerns. NTD. 2026-06-27.
- OpenAI restricts ChatGPT rollout at Washington’s request. RT. 2026-06-27.
- Kevin Hughes. Trump launches Genesis Mission AI initiative compares it to Manhattan Project. NaturalNews.com. 2025-11-30.
- Fable 5 Shutdown, SpaceX Buys Cursor, OpenAI Recruits McKinsey | Weekly Digest. The Creator's AI. 2026-06-19.
- Kevin Hughes. Chinese researchers replicate OpenAI's advanced AI model, sparking global debate on open source and AI security. NaturalNews.com. 2025-01-10.
- OpenAI reveals more details about its agreement with the Pentagon. TechCrunch. 2026-03-01.
- Belle Carter. OpenAI strikes 200M defense pact amid ethical AI debate. NaturalNews.com. 2026-03-04.
- Willow Tohi. Biden gave China the AI advantage. NaturalNews.com. 2025-01-29.
- Willow Tohi. AI arms race or AI suicide pact? A former OpenAI researcher warns of catastrophic risks in unchecked AI development. NaturalNews.com. 2025-02-03.
- OpenAI Limits Access to New Models, Citing Government Security Concerns. WSJ.com. 2026-06-26.
- OpenAI and Anthropic limit new AI models to Trump-approved customers during cybersecurity review. SeattleTimes.com. 2026-06-26.
Please contact us for more information.
