The campaign for the United States and China to sit down and talk about artificial intelligence has run into a flat refusal from one of business television's louder voices.

The Big Money Show panelist Dagen McDowell rejected calls for US-China AI talks during a discussion of Anthropic's Mythos AI and the national security concerns the system raises. Her answer to advocates of diplomacy with Beijing was short and unmistakable: "They can suck it."

The remark is easy to file under cable-news theater. It deserves more weight than that. McDowell's response reflects a view that has been hardening inside American politics for years: AI is a strategic competition, China is an adversary, and there is nothing to gain from pretending otherwise.

The panel's discussion centered on Anthropic's Mythos AI and the national security concerns it raises. That framing matters. Anthropic is one of the leading AI labs, and it has long presented itself as an AI safety company. A lab founded to make advanced AI less dangerous now sits at the center of a security debate where the danger is not the machine but the rival nation building similar machines. That shift has changed the terms of the entire AI policy discussion.

The safety case for dialogue

The strongest argument for US-China AI talks rests on shared risk. The two countries lead the world in advanced AI research, and both are pushing the technology into military planning, logistics, surveillance, and intelligence gathering. AI systems that compress decision-making time on both sides of a confrontation are exactly the kind of technology where a misread move can spiral. Supporters of dialogue argue that the logic behind Cold War hotlines applies here: channels built in calm times are worth more than channels improvised in a crisis, and rivals can communicate without trusting each other.

That case has been made by researchers, former diplomats, and some of the companies building the technology. It carries obvious appeal for a lab like Anthropic, whose stated mission is preventing powerful AI from causing harm, whatever flag flies on the model's code. The more capable AI becomes, the stronger the argument that even adversaries share an interest in avoiding failures that neither side could control.

The national security concern attached to Mythos is not an isolated story. Every major AI lab has seen its work pulled into the geopolitical current, and companies that once emphasized international cooperation now face questions about which side their technology serves. That tension sits at the center of the AI policy fight. The same systems that promise advances in medicine and energy can be directed at military targets, and the line between those uses is drawn by governments, not by developers.

The zero-sum view

The countercase, the one McDowell delivered with such economy, is that China is not a partner in managing shared risk. Its AI industry operates under state direction, and its military and surveillance programs have shown no appetite for the transparency that would make any agreement meaningful. From this view, a summit with Beijing hands the Chinese government a stage, lets it claim good-faith engagement, and produces a piece of paper while the programs that triggered the concern keep moving forward. The phrase McDowell chose was crude. The position behind it is mainstream among national security conservatives, and it is gaining ground.

There is a middle path, and it may be where the argument eventually lands. Washington does not have to choose between a public dialogue that legitimizes China's AI record and a total refusal to communicate. Narrow channels on specific technical risks, clear red lines on autonomous weapons, and quiet warnings about destabilizing deployments could give hawks the toughness they demand and diplomats the communication they say is necessary. The problem with the stance McDowell voiced is that it collapses those options into a single posture. When the answer to any proposal for engagement is a flat refusal, the debate stops being about how to manage China and becomes about whether to bother trying.

McDowell's comment is notable less for its originality than for its directness. She said plainly what a large and growing faction of American politics is thinking. Calls for AI talks with China have multiplied as the technology's strategic weight has grown. So has the backlash. Every new national security scare attached to an AI system, including the one animating this panel's discussion, feeds the conviction that the United States and China are not heading toward a shared safety regime but into a competition with no room for softness.

The phrase "they can suck it" will not appear in any policy paper. It is not the language of the State Department. But it is the language of a public argument that has moved past politeness. For anyone who believes the United States should engage China on AI risks, the comment is a warning that the window for such talks is closing. For anyone who believes competition is the only realistic policy, it is a sign that their side holds the rhetorical ground.

None of this means the calls for talks will go away. The diplomatic machinery in Washington keeps producing proposals for engagement, and parts of the technology industry remain eager for access to Chinese markets. But each public rejection makes the next proposal harder to sell. Politicians who want to be seen as tough on Beijing now have a ready-made vocabulary, and it sounds like the line McDowell drew on The Big Money Show.

Either way, a panel discussion about Anthropic's Mythos AI and national security ended with a sharper question than the one it started with. If the loudest voices in the room have already concluded that dialogue with China is a waste, the real debate moves to what the United States is willing to do instead. That is the discussion that matters, and it is only beginning.