SAN FRANCISCO, California: OpenAI and Anthropic have restricted access to their latest artificial intelligence models following cybersecurity reviews by the Trump administration, marking an unprecedented level of U.S. government scrutiny of advanced AI systems.

ChatGPT maker OpenAI said on June 26 that it is limiting the release of its latest model, GPT-5.6 Sol, to a small group of customers approved by the Trump administration. Hours later, rival Anthropic said the administration had approved a limited release of its cybersecurity-focused model Mythos 5, two weeks after effectively banning it.

Both companies said the models would initially be available only to trusted partners.

"We don't believe this kind of government access process should become the long-term default," OpenAI said in a statement, adding that the testing period is intended as a temporary step on the "path to broader availability in the coming weeks."

Anthropic had earlier withdrawn its Fable 5 and Mythos 5 models after complying with a Trump administration directive restricting their use by foreign nationals. On June 26, the government lifted restrictions on Mythos 5, allowing it to be "redeployed to a small group of cyber defenders and infrastructure providers," Anthropic said.

The White House said it continues to work with leading AI developers to address national security concerns as increasingly powerful AI models emerge.

The heightened oversight follows President Donald Trump's executive order signed earlier in June establishing a framework for the federal government to assess the national security risks of advanced AI systems for up to 30 days before their public release. While participation is described as voluntary, the framework is still being developed.

Officials became increasingly concerned after Anthropic warned earlier this year that its Mythos model was highly effective at identifying software vulnerabilities that malicious hackers could potentially exploit.

OpenAI said its new Sol model "is better at helping people find and fix vulnerabilities" than carrying out cyberattacks and does not exceed the company's internal risk threshold. However, it acknowledged that unforeseen risks could emerge if the model is combined with other tools.

"That uncertainty, along with the model's broader step change in capabilities, is why we are pairing the model's increased capabilities with stronger safeguards and a phased release," the company said.

The government review has drawn criticism from some lawmakers and cybersecurity experts, who argue that the absence of a formal legal process creates uncertainty for AI companies. OpenAI has not disclosed the identities of the roughly 20 customers approved to use GPT-5.6 Sol, while Anthropic said it would continue working with the government to expand access to Mythos 5 and restore broader availability of Fable 5.