The sudden debut of Anthropic’s powerful new Mythos model has sent shockwaves through both Washington and Beijing, prompting quiet discussions this week to explore reviving an emergency artificial intelligence communication channel ahead of President Donald Trump’s state visit to China. The talks represent a significant shift in tone for an administration that had previously taken a laissez-faire approach to international AI governance, underscoring the unprecedented capabilities and risks associated with the latest generation of frontier models.
According to a report by the Los Angeles Times, a senior administration official confirmed on Sunday that the White House is looking to create “a channel of communication for AI like others that they have in many areas that have intense focus with the U.S. and China.” While the exact formality and structure of the channel remain undetermined, the official emphasized the administration’s desire to use the upcoming leaders’ meeting to open a dialogue on the matter.
The Mythos Catalyst
The urgency of these discussions is directly tied to Mythos’s capabilities. Across the industry and within government circles, the model is increasingly viewed not just as a technological marvel, but as a potential cyberweapon of unprecedented power. Experts warn that its advanced reasoning and execution capabilities could theoretically be used to infiltrate and exploit critical digital communication systems, including government databases, financial institutions, and healthcare networks.
This shared alarm has forced a reckoning in both capitals. As we noted in our coverage of the military AI agenda at the Trump-Xi summit, the intersection of advanced AI and national security is becoming impossible to ignore. The potential for a runaway AI scenario or a catastrophic misunderstanding driven by autonomous systems has elevated the issue from theoretical academic debate to immediate geopolitical priority.
Overcoming Past Failures
Reviving the emergency channel will require overcoming significant historical baggage. Previous attempts at establishing AI governance frameworks between the two superpowers have met with limited success. A large U.S. delegation arrived at earlier talks in Geneva with high hopes, only to have those hopes abruptly dashed when the Chinese contingent dismissed American concerns about runaway AI as “academic, almost theoretical.” Instead, Beijing quickly pivoted the conversation to U.S. export controls, viewing them as a thinly veiled effort to stifle China’s technological rise.
Jake Sullivan, the former U.S. national security advisor under President Biden, highlighted this dynamic in an interview with the Los Angeles Times. “They naturally view any American diplomatic initiative involving limitations or restrictions of one flavor or another on a capability as being a trap,” Sullivan noted.
Despite this distrust, a breakthrough was achieved in November 2024 in Lima, Peru, where both sides agreed to keep AI out of the command and control of nuclear weapons. Sullivan described this as “a breaking of the seal that we could actually do something on AI.” However, he noted that the incoming Trump administration initially showed little interest in continuing the dialogue, preferring a more hands-off approach. “That’s all changed in the past few weeks,” Sullivan added, pointing to the shifting landscape .
Divergent Strategies and the AGI Race
Any formal discussions on AI governance will be complicated by the fundamentally different strategies pursued by the United States and China. The U.S. ecosystem, driven by tech giants like OpenAI, Google, Anthropic, Microsoft, and Meta, is locked in a winner-takes-all race to achieve Artificial General Intelligence (AGI). The prevailing theory is that the first to reach AGI will secure an insurmountable advantage through recursive self-improvement.
In contrast, China has adopted a state-sanctioned strategy focused on the diffusion and application of AI across specific industries. As Scott Kennedy, a senior advisor at the Center for Strategic & International Studies (CSIS), explained, “The Chinese believe there is no single race, but multiple races. The U.S. is focused on achieving AGI, while China is focused on diffusion and applications of AI into the rest of their economy—manufacturing, humanoid robotics, all aspects of the internet of things”. This aligns with China’s broader industrial goals, such as the deployment of humanoid robots in manufacturing.
This divergence in focus creates a complex negotiating environment. If the U.S. is the primary driver toward AGI, Chinese negotiators may argue that the responsibility for mitigating its existential risks lies squarely with Washington. Furthermore, the massive funding gap between U.S. and Chinese AI firms—with OpenAI valued near $800 billion while leading Chinese public companies hover around $20 billion—adds another layer of tension. Matt Sheehan, a senior fellow at the Carnegie Endowment for International Peace, noted this disparity, emphasizing that “It’s not just chips. It’s money” .
The Path Forward
Despite these challenges, there are signs that Beijing recognizes the gravity of the situation. Daniel Remler, who led AI policy at the State Department during the Biden administration, cast doubt on claims that China is uninterested in AGI risks. “Chinese technologists are close observers of the U.S. AI ecosystem, and sometimes they say what they think,” he noted. “Many were impressed by the [Mythos] model to the point of despair”.
As President Trump prepares for his state visit, the establishment of an emergency AI channel represents a critical test of whether the U.S. and China can compartmentalize their broader strategic competition to address shared existential risks. While a comprehensive arms control agreement for AI remains a distant prospect, a dedicated line of communication could provide a vital safety valve in an increasingly volatile technological landscape. The outcome of these talks will likely shape the trajectory of global AI governance for years to come.