OpenAI and Anthropic have limited public access to their latest artificial intelligence models as the Trump administration expands its review of advanced AI systems for potential cybersecurity and national security risks.
OpenAI announced Friday that its newest model, GPT-5.6 Sol, will initially be available only to a select group of customers approved through a federal review process. The company described the arrangement as a temporary measure while government officials complete evaluations of the model’s security implications.
“We don’t believe this kind of government access process should become the long-term default,” OpenAI said in a statement, adding that it expects broader availability in the coming weeks after the testing period concludes.
The announcement came hours after Anthropic said federal officials had approved a limited return of its advanced cybersecurity model, Mythos 5, following restrictions imposed earlier this month by the U.S. Commerce Department.
Government Review Expands for Frontier AI Systems
The White House said it continues to work with leading AI developers to address the risks associated with increasingly capable artificial intelligence technologies.
The administration’s heightened scrutiny follows concerns raised earlier this year after Anthropic disclosed that its Mythos model demonstrated advanced capabilities for identifying software vulnerabilities. Officials have expressed concern that such capabilities could potentially be exploited by malicious actors targeting critical infrastructure.
Earlier in June, President Donald Trump signed an executive order establishing a framework allowing the federal government to evaluate the national security implications of the most advanced AI systems for up to 30 days before broader public deployment. Although participation is described as voluntary, the review process remains under development.
Anthropic Receives Partial Approval
Anthropic had temporarily withdrawn two recently introduced AI models—Fable 5 and Mythos 5—after complying with a federal directive restricting access by foreign nationals.
On Friday, the company said the government had authorized a limited redeployment of Mythos 5 to selected cybersecurity organizations and infrastructure providers while broader restrictions remain in place.
Anthropic said it welcomed the partial approval and intends to continue working with government officials to expand access to Mythos and restore availability of Fable.
A letter from U.S. Commerce Secretary Howard Lutnick, dated Friday, stated that Anthropic’s efforts to address government concerns had produced “significant progress.”
OpenAI Emphasizes Gradual Rollout
OpenAI said its GPT-5.6 Sol model is designed to improve the identification and remediation of software vulnerabilities while maintaining safeguards intended to reduce offensive cyber capabilities.
The company acknowledged that powerful AI systems can present unforeseen risks, particularly when combined with other advanced technologies, and said those uncertainties influenced its decision to adopt a phased release strategy with enhanced security measures.
OpenAI has not identified the approximately 20 organizations that have been approved to use the model during the initial rollout.
Critics Raise Concerns Over Government Control
The federal review process has drawn criticism from lawmakers and cybersecurity experts who argue that the absence of a formal legal framework could create uncertainty for AI developers.
Representative Lori Trahan of Massachusetts said she is concerned that executive branch officials are determining access to advanced AI systems without clear statutory authority or transparent oversight procedures.
Cybersecurity researcher Alex Stamos also questioned the government’s earlier decision to restrict Anthropic’s Fable model, arguing that available research did not identify risks substantially different from those associated with other advanced AI systems already available globally.
Stamos warned that limiting domestic AI innovation could weaken the United States’ competitive position in the global technology race.
AI Oversight Coincides With Corporate Growth Plans
The expanding federal review process comes as both OpenAI and Anthropic continue exploring potential public offerings and broader capital market opportunities.
OpenAI Chief Executive Sam Altman recently met with Commerce Secretary Howard Lutnick to discuss the release of GPT-5.6 Sol as part of ongoing discussions between government officials and AI industry leaders.
Anthropic has also participated in those discussions while continuing separate legal proceedings involving federal restrictions on the use of its Claude chatbot by government agencies.
President Trump has also floated the possibility of the U.S. government holding ownership stakes in leading artificial intelligence companies, suggesting a model in which the American public could become a partner in the industry’s future growth.
