The United States is actively managing the release of advanced artificial intelligence models, a strategy that has led to restrictions on domestic firms like OpenAI and Anthropic. This comes as U.S. officials express concerns about China's progress in the artificial intelligence race, particularly in cybersecurity AI capabilities.
Beijing-based tech firms 360 Security Technology and Z.ai (Zhipu AI) have significantly narrowed the gap with U.S. competitors in identifying software vulnerabilities, according to Zhou Hongyi, CEO of 360 Security Technology. He stated at a cybersecurity conference that his company's bug-finding tool, Tulongfeng, is now comparable to Anthropic's Mythos, a tool designed for similar functions. Zhou remarked that such powerful tools for cyberwarfare should not be exclusively in American hands.
The Trump administration has implemented measures to vet new AI models due to security concerns. OpenAI announced it was restricting the release of its new AI model, GPT-5.6 Sol, at the request of the administration. Similarly, Anthropic received approval for a limited release of its cybersecurity model, Mythos 5, after an initial ban by the Commerce Department. Both companies are making these powerful models available only to small groups of vetted partners. OpenAI has described this staggered release as a temporary measure before broader availability.
Earlier in June, President Trump signed an executive order establishing a framework for the federal government to assess national security risks associated with advanced AI systems prior to their public release. Participation by AI developers is described as voluntary, and the framework is still under development.
Jacob Helberg, the deputy secretary of state for economic affairs and a former tech advisor, indicated that the administration is closely monitoring Chinese open-source AI models. He noted that while China is closing the gap in some areas, it still trails behind on others, such as OpenAI's products. Concerns have been heightened by warnings from companies like Anthropic, whose Mythos model was identified as being capable of finding software flaws that could be exploited by malicious actors. Investor David Sacks, who advises Trump on technology and science, described Mythos as a "cyber weapon."
Despite these technological rivalries, President Trump and Chinese President Xi Jinping have maintained what Trump describes as a "good relationship." However, the broader geopolitical tension between the two superpowers persists.
U.S. Representative Lori Trahan has voiced concerns about the administration's case-by-case approach to granting access to new AI models, stating that decisions appear to be made by appointees without a clear legal process or oversight.
Meanwhile, the U.S. has previously cleared approximately ten Chinese firms to acquire Nvidia's H200 AI chip, the company's second-most powerful offering.