Artificial Intelligence & Machine Learning
,
Governance & Risk Management
,
Government
Delayed Rollout Highlights Enterprise Risk as Frontier AI Becomes Strategic Infrastructure
The Trump administration has lifted restrictions on OpenAI’s GPT-5.6 model family, clearing the way for a broad public release on Thursday, Axios reported.
See Also: Edge Transformation: Top 5 SASE Predictions and Trends
The decision comes as officials and artificial intelligence developers are still grappling with how much oversight the U.S. government should have before powerful new AI models are made broadly available, especially when models have demonstrated advanced cybersecurity, coding and scientific capabilities.
OpenAI announced Tuesday evening on X that GPT-5.6 – including its flagship Sol model and lower-tier Terra and Luna models – will become publicly available on Thursday, July 9. The company has said the model family advances capabilities in software engineering, computer use, professional knowledge work, scientific research and cybersecurity.
OpenAI said GPT-5.6 Sol is its most capable cybersecurity model to date and improves performance on long-horizon security tasks, including vulnerability research and exploitation. The company said Sol is better at helping users find and fix vulnerabilities than at reliably carrying out end-to-end attacks, but acknowledged that benchmarks can’t capture every way a model may be used or combined with other tools.
The pending release follows additional testing and meetings between OpenAI and federal officials, Axios reported.
Testing was conducted by the Center for AI Standards and Innovation within the U.S. Department of Commerce, and OpenAI technical experts have remained in Washington to address potential government questions, Axios reported. The Trump administration pushed OpenAI last month to conduct a staggered release of GPT-5.6, initially limiting access to government-approved entities.
The White House, however, disputed the idea that it had formally approved or “given a green light” to the model. “No such permission is required or granted,” the official said in a statement to Axios, adding that the decisions on the timing and scope of a model release “rest entirely with the companies.”
The official also pointed to Trump’s June 2 executive order, which Axios said bars mandatory federal licensing or preclearance for AI model releases, and said any testing or meetings with government experts are voluntary.
For its part, OpenAI previously said the limited preview was not its preferred long-term approach, Axios reported. In a June 26 announcement, the company said it was beginning with a small group of trusted partners after engaging with the U.S. government, but said such a government access process should not become the default because it delays access for users, developers, enterprises, cyber defenders and global partners.
The government’s involvement in the GPT-5.6 rollout shows that frontier models increasingly stand apart from ordinary commercial software products when it comes to regulation and national security. Last month, after the U.S. Department of Commerce invoked export controls on Anthropic’s Fable 5 and Mythos 5 models, the frontier AI company took its products off the market. The government withdrew the controls after three weeks of negotiations with Anthropic over potential security vulnerabilities.
“Frontier AI has become a matter of national strategy,” said Jim Sherlock, AI practice lead at cybersecurity consulting firm ProCircular. “Governments are increasingly treating these models less like software products and more like strategic infrastructure because of their potential economic, intelligence and cybersecurity impact.”
For enterprises, the reported end of restrictions should not be treated as a security endorsement, Sherlock said. “It doesn’t validate how safely a model can be integrated into your business, connected to sensitive data, or embedded into autonomous workflows. Those responsibilities remain with the organization.”
AI model availability is an operational risk for businesses, he said, and as governments get more involved in regulating advanced AI, organizations need to assume that access could change at any time due to geopolitical events, export controls or evolving policy.
“The companies that will be best positioned aren’t those betting everything on one model,” Sherlock said. “They’re the ones building flexible AI architectures that can adapt as the landscape changes.”