Skip to content 
NOYou can now listen to Fox News articles!
America and China do not agree on much. But when it comes to AI, we all lose when we lose control.
Last month, as ranking member of the House Select Committee on China, I convened the first congressional hearing to call for a binding, enforceable, and verifiable US-China AI treaty. I also led every Democrat on the committee to send a letter to the White House demanding global action to address AI risks. Within days, President Trump and General Secretary Xi announced they would set up a bilateral dialogue on AI. This is an important first step that would not have happened without congressional pressure.
But this moment requires more than discussion. At my hearing, some of the world’s leading experts on AI called on the United States and China to form joint technical working groups with the top scientists of both countries. Their mission would be to develop mutually trustworthy verification technologies that would prevent both sides from cheating when signing a US-China AI treaty. This is a simple second step that we should take today.
ANTHROPIC CEO LIKES AI FIGHT WITH CHINA IN COLD WAR, SEEKS ‘ODERARMAMENT NEGOTIATIONS’
Critically, we also need to imagine what an actual treaty to save humanity from uncontrolled AI would look like. The risk that humans will lose control over AI or that misaligned models could unleash biological havoc is increasing every day, and if the world’s AI superpowers don’t act soon, it may be too late.
To address these risks, there are five key pillars of a US-China AI agreement. First, we need a worldwide break on recursive self-improvement. This is when AI develops AI without human intervention, and it is the most likely way for humanity to completely lose control of this technology. Asking AI to design its own security measures is like asking the fox to guard the henhouse.
MIKE JOHNSON REJECTS MORATORIUM ON CHINA BECAUSE CONSERVATIVE ACTIVIST CALLS FOR GUARDRAILS
Second, we should impose a global moratorium on superintelligence. As it has traditionally been defined, superintelligence is AI that exceeds human capabilities across any domain. This could open Pandora’s box, and neither America nor China is ready for it.
Third, we must have kill switches embedded in all advanced models. While scientists disagree on the effectiveness of kill switches, they would at least give us a fighting chance if rogue models decide to turn against humanity.
AI KILL SWITCH BILL CAN SHUT DOWN ROGUE MODELS
Fourth, every AI model should be tested in an air-gapped sandbox before being released to avoid catastrophic misuse. Whether it is an American model or a Chinese model, criminals, terrorists and even misaligned AI agents should not be able to unleash economically devastating cyber attacks because a border lab decided to release a beta version.
And fifth, but perhaps most important, we must heed President Reagan’s wise words and “Trust, but verify.” Especially with China, we cannot simply take the CCP at its word. Both governments should agree on verification technologies and inspection regimes that ensure Beijing cannot simply cheat, allowing its labs to run ahead while American labs lag behind. We also need to build institutions that both sides can trust in a fierce competition, such as an International AI Inspection Agency.
BEYOND KILLER ROBOTS: China’s REAL AI Threat to America is ALREADY HERE, EXPERTS WARN
During the Cold War, even as America and the Soviet Union fought fiercely, President Reagan and Soviet leader Mikhail Gorbachev recognized that nuclear war would have left both nations worse off. Winning an AI war is like winning a nuclear war. It is catastrophic for civilization.
Even as we continue to hold China accountable for its predatory economic practices that harm American workers and its human rights abuses, both countries should be able to rally around the common cause of keeping humans in control of AI. (Kenny Holston/Pool Photo via AP)
In the same spirit, even as we continue to hold China accountable for its predatory economic practices that harm American workers and its human rights abuses, both countries should be able to rally around the common cause of keeping humans in control of AI. A US-China AI treaty benefits neither America nor China, but rather all of humanity. A global pandemic unleashed by rogue AI would not distinguish between American and Chinese citizens, and neither government would benefit if humanity stopped calling the shots.
CLICK HERE FOR MORE FOX NEWS OPINION
A verifiable US-China AI treaty won’t happen overnight, but if we don’t start the hard work of negotiating such an agreement soon, humanity will pay a steep price. In the meantime, in addition to the formation of joint technical working groups with China, we must also get our own house in order.
For Congress, this means passing laws that prevent AI companies from putting profits over the collective destiny of humanity. For American AI laboratories, this means investing in security and strengthening their cybersecurity to prevent China from stealing our edge. If China’s lax cybersecurity allows us to illegally obtain advanced model weights, it not only harms our national security, but also undermines our strongest leverage in pursuing a US-China AI treaty.
CLICK HERE TO DOWNLOAD THE FOX NEWS APP
When it comes to catastrophic AI risks, it cannot be America versus China or Republicans versus Democrats. The arrival of superintelligence arguably marks a more consequential moment for our species than even the dawn of the Industrial Revolution or the development of the atomic bomb. While the former has transformed society and the latter risks its utter destruction, advanced forms of AI threaten to do both. Let’s come together and keep humanity in the driver’s seat.
CLICK HERE TO READ MORE FROM REP. RO KHANNA
