OpenAI and Anthropic Restrict New AI Models to Trump-Approved Partners Amid Cybersecurity Oversight

In a significant development within the artificial intelligence landscape, OpenAI announced on Friday that it will limit the release of its latest AI model, GPT-5.6 Sol, at the request of President Donald Trump’s administration. This move marks a notable instance of government intervention in the AI sector, particularly concerning cybersecurity risks associated with advanced technologies.

Shortly after OpenAI’s announcement, its primary competitor, Anthropic, revealed that the Trump administration has approved a limited release of its advanced cybersecurity model, Mythos 5. This approval comes two weeks after the U.S. Commerce Department had effectively banned the model’s deployment.

Both companies have stated that their new models will only be accessible to a select group of trusted partners. OpenAI specified that its GPT-5.6 Sol would be available exclusively to customers authorized by the Trump administration. The company expressed its concerns regarding the long-term implications of such government oversight, stating, “We don’t believe this kind of government access process should become the long-term default.” OpenAI views this testing phase as a temporary measure, aiming for broader availability in the near future.

Government Scrutiny of AI Models

The staggered release of OpenAI’s new AI system follows recent government actions against Anthropic, which had to take its newly unveiled models, Fable 5 and Mythos 5, offline to comply with a directive from the Trump administration that barred their use by foreign nationals. On Friday, the government lifted restrictions on Mythos 5, allowing it to be redeployed to a limited group of cyber defenders and infrastructure providers.

The White House has indicated that it continues to collaborate with leading AI laboratories to tackle the challenges posed by rapidly advancing technologies. Concerns have escalated since Anthropic warned that its Mythos model could identify software vulnerabilities that might be exploited by malicious actors, potentially threatening critical computer networks globally.

In June, Trump signed an executive order establishing a framework for the federal government to assess the national security risks of advanced AI systems before their public release. This framework allows for a vetting period of up to 30 days, although participation from AI developers remains voluntary and the specifics of the framework are still under development.

Industry Reactions and Concerns

Some allies of Trump have attributed the heightened scrutiny to Anthropic and its CEO, Dario Amodei. Investor David Sacks, who co-leads Trump’s council of technology and science advisers, remarked that Amodei’s claims about creating a cyber weapon called Mythos heightened concerns among officials.

OpenAI, based in San Francisco, has described its new Sol model as being more effective at identifying and rectifying vulnerabilities than executing cyberattacks. However, the company acknowledged the potential for unforeseen risks, particularly if the model is integrated 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 stated.

As of now, OpenAI has not disclosed the identities of the approximately 20 customers approved to utilize the new model.

Criticism of Government Intervention

U.S. Representative Lori Trahan, a Democrat from Massachusetts and co-author of a bipartisan bill aimed at regulating AI, expressed her concerns regarding the Trump administration’s selective approval process for accessing new AI models. She stated, “No law. No process. No oversight. Just appointees in Washington deciding who’s in and who’s out.”

A coalition of technology experts has also criticized the government’s actions that led to the shutdown of Anthropic’s Fable model, which was presented as a safer alternative to Mythos. Despite the lifting of restrictions on Mythos, Fable remains unavailable.

Stanford University cybersecurity expert Alex Stamos voiced skepticism regarding the government’s rationale for its actions. He noted that an analysis conducted by Amazon, Anthropic’s primary cloud computing partner, found no risks associated with Fable that are not present in other publicly available AI models, including those developed in China. Stamos remarked, “If the administration is honest about wanting the United States to beat China in this race, then this is about the dumbest thing they could possibly do.”

Increased Oversight Amid IPO Aspirations

OpenAI CEO Sam Altman recently engaged in discussions with U.S. Commerce Secretary Howard Lutnick regarding the model release, part of ongoing negotiations between AI industry leaders and Trump officials. Anthropic has also participated in these talks, although Amodei’s relationship with the Trump administration has been more contentious. The Pentagon has classified Anthropic as a national security risk due to its ethical and safety concerns regarding AI applications in warfare, leading to a directive for federal agencies to cease using its Claude model. Anthropic has since filed a lawsuit, which is currently progressing through federal courts.

On Friday, Anthropic expressed satisfaction with the partial release of Mythos and indicated its intention to collaborate with the government to expand access and reintroduce Fable to general users. Lutnick informed Anthropic in a letter that their efforts to address governmental concerns had resulted in “significant progress.”

The intensified scrutiny from the government complicates the efforts of both OpenAI and Anthropic as they explore options for going public, particularly following SpaceX’s record-setting initial public offering on June 12.

Trump has suggested the possibility of the U.S. government acquiring stakes in leading AI companies, proposing a model where “pieces could be given to the American public, where the American public essentially becomes a partner with the companies.”

For further information, visit the original reporting source: SecurityWeek.

Keep reading for the latest cybersecurity developments, threat intelligence and breaking updates from across the Middle East.