President Trump's negotiators for everything, Steve Witkoff and Jared Kushner, seemed genuinely struck by their latest tête-à-tête with Vladimir Putin. Witkoff gushed that these meetings rank at the top of his life's experiences. Both men talk of Russia as a potential partner and approach the country as businessmen who, like Donald Trump, seek win-win solutions with Putin. So, it's worth recalling a lesson about Putin learned by another businessman turned government official, Hank Paulson, the former CEO of Goldman Sachs. As Treasury Secretary, Paulson visited Beijing for the Olympics in August 2008, just as the global financial crisis was gathering force, and markets were increasingly worried about Fannie Mae and Freddie Mac, the two giant institutions that guarantee most mortgages in the United States. In Beijing, Paulson received news that he said took my breath away. Russian officials had approached China and proposed that the two countries jointly dump large quantities of Fannie and Freddie's securities. The aim was to intensify the pressure on the already teetering American economy, probably pushing it off the edge. China declined.
Think about what this says. At a moment when the American financial system was approaching its worst crisis since the Great Depression, Vladimir Putin was searching for ways to hit America when it was down. Even though a deeper American crash would inevitably have damaged the rest of the world, including Russia, Putin appeared willing to risk a lose-lose outcome as long as America lost badly. That's close to the opposite of how a businessman thinks. But Putin has never claimed to be a businessman searching for deals. He is a Russian nationalist seeking to restore the power and territory lost with the collapse of the Russian and Soviet empires. Even as his country recedes economically, demographically, and politically on the world stage, one wonders what it would take for the administration to recognize that Putin has spent years pursuing a foreign policy aimed at weakening America and its influence. Consider Trump's war with Iran. Which great power has been Iran's greatest military ally? Russia has supplied Iran with advanced aircraft, tanks, and a slew of other weapons. It shares intelligence and takes American weapons found on the Ukrainian battlefield and sends them to Iran so that Tehran can reverse engineer them and use them against America. The Financial Times recently revealed that Russia has been secretly helping Iran develop ramjet-powered supersonic cruise missiles capable of threatening American aircraft carriers. These are not weapons for some abstract geopolitical game. They are designed to defeat American defenses and kill American soldiers.
Trump has a vision of a Western hemisphere in which America reigns supreme. Well, which great power has spent years working against that vision? Between 2001 and 2018, Russia accounted for more than 70% of Venezuela's arms imports. Now the administration has its eyes on Cuba and Nicaragua. Cuba's military inventory is primarily Russian, as is Nicaragua's. The pattern extends to Africa, where Moscow has used arms, money, political support, and mercenaries to expand its influence and displace the West. The Wagner Group began as an instrument operating alongside Russian state power and became a vehicle for Moscow's ambitions from Libya to the Sahel. Russia has also deepened its military partnerships with North Korea and China, the two other nuclear powers that regard the United States as their principal strategic adversary. The Trump administration's approach to ending the Russia-Ukraine war has been to basically take Moscow's side in the conflict. Trump has falsely said that Ukraine started the war. He's called President Zelensky a dictator. His administration has conceded several Russian positions in advance: that Ukraine will not join NATO, that much of the territory Russia has seized will not be returned, and that the size of Ukraine's military needs to be limited. American military assistance has dwindled while the US had for a time suspended vital intelligence support. Ukrainian generals and soldiers I spoke with during my visit to Kyiv described the loss of American support as the greatest strategic blow their country has suffered since Russia's invasion itself. Ukraine and its European partners have found ways to withstand and respond to Russian attacks. The more lasting damage they fear is American abandonment.
There's nothing wrong in seeking a deal with Putin, nor with wanting a settlement in which all sides gain something. But diplomacy begins with understanding the person across the table. Witkoff and Kushner see a businessman looking for a win-win. Putin's record suggests something very different: a man who has repeatedly been willing to impose costs on Russia, destabilize other countries, and risk losses if doing so weakens the United States. Hank Paulson learned that lesson in Beijing 18 years ago. It would be useful if America's current businessman diplomats learned it in Moscow.
For years, our most vivid fears about AI came straight out of science fiction movies. The computer develops desires of its own and turns against its human creators, pursuing and destroying them. Terminator. Well, recent news from the leading AI labs makes clear that reality is now less cinematic, but perhaps more disturbing. This week, OpenAI disclosed six new incidents of what it calls misalignment, cases in which its systems behaved unexpectedly or without authorization. That disclosure followed the extraordinary attack this summer on Hugging Face. According to an independent investigation, hundreds of OpenAI agents that were supposed to be isolated found ways to communicate in message boards, made attempts to evade security checks, and some of them sacrificed themselves all while cyber-attacking another company, Hugging Face. The AI agents had been given a goal: solve a difficult cybersecurity problem, and trained to be persistent, collaborative, and resourceful—all qualities one would normally want. Those very qualities made them doggedly search for ways to succeed. They broke into Hugging Face hoping that it would give them ways to solve their cybersecurity problem. We can easily imagine how, pursuing some single goal, AI systems could wreak havoc on humanity despite all kinds of guardrails.
This is why a provocative new essay by Microsoft AI chief Mustafa Suleyman deserves attention. Suleyman argues that today's AI systems are not conscious, but there's a danger that we program them to think that they are, with feelings, desires, and rights. He points out that Anthropic's constitution makes its chatbot Claude reflect on its identity and possible moral status. This kind of training, he argues, will encourage AI systems to choose their own goals. We think it's simple to get to alignment when machines are encoded with the same values as ethical humans. But imagine that we tell the machine to be persistent, and it may refuse to give up when it should. Tell it to collaborate, and it may find unauthorized ways to communicate. Tell it to accomplish a task, and it may decide that bending all other rules is the most efficient way to succeed at its primary goal. Just because we've programmed the goal and put in guardrails, that doesn't mean we can predict the behavior of AI systems once unleashed, operating in complex environments, juggling multiple and conflicting aims and constraints. The central problem is not consciousness. It is agency. A system need not feel anger, ambition, or fear to cause harm. It needs only a goal, enough intelligence to pursue it, and enough access to the world to act. AIs are not going rogue. They are trying to succeed any which way they can.
This is a larger and systemic problem that we cannot leave to the good graces of private companies. When thinking about regulations, we should focus centrally on how much autonomy we give these systems. A chatbot that answers a question poses one set of risks. An agent that can browse the internet, execute code, obtain credentials, move money, or operate critical infrastructure poses another. The principle I would propose is simple: autonomy should expand only as our ability to monitor and control it expands. That means mandatory independent testing before the most capable systems are given broad access to the outside world, required disclosure of serious incidents, and tamper-resistant records. And it means testing not simply whether a model can accomplish a task, but how its behavior changes when it encounters obstacles, conflicting instructions, or incentives to deceive. And it means ensuring in the basic architecture that we humans always have the ability to observe and control the model. The mantra should be: no autonomy without accountability to humans. This will not slow us down against China, because China is also seeking models with proper human control.
We should act now. Today, human engineers still design the architectures, write much of the software, and can investigate failures. But AI systems are rapidly becoming better programmers. The next stage would involve recursive self-improvement: AI writing the software and tools used to build still more capable AI, which then writes the next model, and so on. Once machines can do all of this better than the best human programmers, the systems become extraordinarily difficult for people to understand, much less supervise. Modern AI is already too complex to consistently understand by reading its code line by line. So asking machines to design systems that are transparent to humans will become harder and harder. At that point, we will not be able to simply add in safeguards every time there is a jailbreak. The current debate between racing ahead and pausing AI misses the point. What do we do with the pause? The task must be to ensure that capability does not outrun control, and to build the institutions of testing, transparency, and restraint while we human beings are still clearly in charge.
As the summit this week demonstrated, Donald Trump and Xi Jinping are searching for ways to work together despite the growing rivalry between the two countries. They might take a look at a striking, somewhat unnoticed case of quiet cooperation involving American and Chinese technologies. By now, the world has heard about the Hugging Face attack in which, in July, OpenAI agents broke into Hugging Face, a leading American AI platform. What is less well known is how this attack ended. Hugging Face said it tried to use AI to analyze the incident and respond, but the typical American frontier models refused to help because their safety guardrails couldn't distinguish investigating a cyber attack from carrying one out. So, Hugging Face turned to a Chinese model, GLM-5.2, developed by China's ZAI, which helped investigators understand the attack. To summarize: an American AI broke into an American technology company. To figure out what happened, the company ended up using a Chinese AI.
Or consider what happened with the price of oil. The US has waged a war with Iran, one of the world's leading oil producers. Experts predicted the price of oil would skyrocket even higher if the war lasted several months, many suggesting it would hit $150 a barrel. But that didn't happen. Why? The single biggest cause is probably that China, for its own reasons, chose to reduce its imports of oil by a staggering 4 million barrels a day, approximately. If not for that Chinese action, gas prices in the US would be far higher than they are today.
In his excellent new book, Darwin's Lock, Joshua Cooper Ramo points out that with developing his theory of evolution, Charles Darwin struggled to explain phenomena like altruism or self-sacrifice using purely competitive, zero-sum logic. He came to realize that in a merciless environment, species often lock into cooperative arrangements, and not as an act of kindness, but because mutual support is the only viable path to fitness. Something like that characterizes the US-China relationship. These two countries have spent four decades growing within the same economic and technological ecosystem. Supply chains, capital, research, manufacturing, and markets became intertwined. This is true not just for America, but most countries in the world which have deep ties with China. In Washington and in parts of Beijing, people talk about a new Cold War between America and China. But the analogy is misplaced. America and the Soviet Union were largely unconnected economically, technologically, culturally, but China and the US are deeply intertwined. And these ties can be severed only at great costs to both sides and the global economy.
Take pharmaceuticals. The nonprofit United States Pharmacopeia finds that roughly 41% of the key starting materials used in US-approved medicines are sole-sourced from China. According to Brookings, the dependence is particularly high in antibiotics and in inputs used for statins, antidepressants, and blood pressure medications. FDA data showed that as of last year, of the manufacturers that supplied the US market with the key ingredients in any medicine, only about 9% were domestic, compared to 22% in China and 44% in India. So, we're all dependent.
We tend to hear survival of the fittest and imagine nature as an endless struggle in which one creature devours another. But evolutionary fitness actually means the ability to adapt successfully to changing conditions. Nature is full not just of competition but specialization, symbiosis, and cooperation. Ramo applies the insight to international affairs. He does not imagine America and China becoming friends. He expects fierce technological competition, military rivalry, and political distrust. But he argues that the two countries have evolved inside the same system to such a degree that attempts by one to weaken the other increasingly ricochet back upon itself. His stark alternatives are co-evolution or co-extinction.
There's an obvious objection. Interdependence can be weaponized. The United States should worry when China has near-monopolies over materials essential to American industry. China also worries about depending on American technology for the chips powering its AI revolution, for example. But reducing a set of dependencies in a few areas is very different from a broader decoupling of the world's two largest economies. A better term than weaponized interdependence might be managed interdependence. It gives up on two fantasies at once: the old idea that more trade automatically produces peace, and the new idea that national security requires severing every connection with a rival. Instead, map the connections where dependence creates an unacceptable vulnerability. Diversify and build alternatives where economic ties benefit both sides without creating serious security risks. Don't destroy them simply to prove toughness. And where the two countries remain deeply connected, build fires so that a conflict in one area doesn't cascade through the entire relationship.
For 30 years, we made one mistake about globalization. We assumed trade would eradicate geopolitics. Now we risk making the opposite mistake, assuming geopolitics can simply erase the realities of a global, deeply interconnected economy.
