Dagen McDowell Rejects Calls for AI Talks with China, Calls for Unleashing U.S. Models

Panelists debated national security implications of advanced AI, focusing on Anthropic’s Mythos model amid concerns over vulnerabilities and potential threats from China.

Dagen McDowell strongly opposed engaging China in discussions about AI safeguards. “They can suck it and catch up which they never will if we do not screw it up here at home,” McDowell said, arguing against over-regulation of technology that policymakers do not fully understand.

McDowell advocated aggressively deploying AI across U.S. national security infrastructure to identify and address vulnerabilities. “These models, we should unleash them across our entire national security infrastructure. That way, we are protected,” McDowell stated. She described Mythos as a tool capable of scouring every nook and cranny of systems to root out weaknesses, emphasizing that such models developed by American companies acting in the best interest of the United States represent strengths, not threats.

The discussion highlighted cybersecurity risks and client inquiries directed at experts about defending against AI-driven attacks. One panelist noted that companies in cybersecurity are actively seeking ways to update systems and build defenses, viewing heightened awareness as a positive development. McDowell reinforced that the primary concern is not American AI firms like Anthropic but “Communist China,” drawing agreement from at least one other panelist.

While supportive of domestic innovation, other panelists expressed caution. They referenced past instances of U.S. companies outsourcing jobs to China and questioned whether technology firms might face similar vulnerabilities, particularly given China’s history of intellectual property theft. Concerns were raised about the large number of Chinese nationals earning degrees in computer science and related fields, and potential differences in how China might approach AI with a more “cynical” mindset compared to American optimism.

One participant compared AI development to the early COVID-19 response, arguing against suppression in favor of controlled release and adaptation. “Put it out there, control it yourself. Know what you are dealing with versus trying to suppress it and you will be able to move forward faster and better,” the panelist said.

The panel acknowledged the challenges of determining appropriate regulation levels, describing upcoming debates as a “dogfight,” but several voices favored unleashing AI capabilities—such as for fraud detection in banking—while maintaining vigilance on national security and foreign adversaries. The conversation is set to continue.