The Diplomat author Mercy Kuo regularly engages subject-matter experts, policy practitioners, and strategic thinkers across the globe for their diverse insights into U.S. Asia policy. This conversation with Dr. Matthew Botvinick  – author of “AI and Political Reform: The Risk to Democracy and How to Respond” (Princeton University Press 2026), world-renowned AI researcher and neuroscientist, senior fellow at Yale Law School and member of the technical staff at Anthropic, leading research on AI and the rule of law – is the 529th in “The Trans-Pacific View Insight Series.” 

Explain how AI will pose fundamental challenges to constitutional democracy.

We are in a remarkable political moment right now, where two trends are converging. First, AI is generating feelings of insecurity and vulnerability across society. Second, in many democracies, trust in government institutions is low – in the U.S., trust is at historic lows. This is a combustible combination, because it provides ideal conditions for what has been called “the demagogue’s playbook,” where populist sentiment is exploited by political leaders to undermine democracy. 

Democracy has always been vulnerable to this risk. The problem now, though, is that AI dramatically raises the stakes. As others have noted, if an authoritarian government can get a foothold, AI makes it much easier for such a government to entrench itself, through surveillance and control of the information citizens can access or share. So the problem is two-fold: AI is creating conditions that can be exploited by aspiring authoritarian leaders, and it is also making available tools that such leaders might use to achieve their goals.  

It is important to add that all of this is about risks, not certainties. In fact, if democracies rise to meet the moment, AI might prove to be a benefit to democracy, rather than a detriment. The first order of business is to get our institutional houses in order, instituting government reforms – in emergency law, surveillance law, separation of powers, and free speech doctrines – that increase democratic resilience. If we can do that and make it past the immediate challenge, then we can start thinking about how AI might be used to empower and inform citizens, support democratic deliberation, and enhance government transparency and accountability. If we are careful and strategic, AI may be a boon rather than a curse. 

Examine the dynamics between rising powerful AI systems and eroding democratic institutions and implications for political and economic freedom.

Democracy depends on a balance between the power of government and the power of society. What keeps that balance in place is a combination of two things. On the government side, to keep power in check, it must be subdivided. Things must be structured to prevent perfectly coordinated action, which would make abuse of power too easy. This is why we have separation of powers and traditional checks and balances. On the societal side, in contrast, adequate power depends on factors that support coordinated action – the news media, the activities of civil society, a free marketplace of ideas, free and fair elections, and so forth. 

All of this means that democracies face danger when (1) the checks and balances within government erode, and (2) when the ability of society to engage in collective action is diminished. 

If we are not careful, AI will threaten to do both of those things. My book unpacks this in a number of ways but let me give just two examples. On the government side, AI might over-empower the executive branch relative to legislative and judicial branches, if AI is used to automate perfect obedience to central executive authority. On the society side, the ability of citizens to act in concert could be undermined by AI-enhanced surveillance or propaganda. Society’s power decreases, government power increases, and the balance required for democracy is lost.

Again, these are risks, not inevitabilities. If we act wisely, AI could enhance society’s capacity for collective action, and keep government’s power in balance, through exactly the means I mentioned earlier. But that won’t happen without strategic and proactive interventions. 

Analyze how the propagation of AI fuels authoritarianism in China and the United States.

The situation in China stands as a warning of what is possible. China’s use of AI to monitor and control its people has been detailed in excellent books like Josh Chin’s and Liza Lin’s “Surveillance State.” The importance of AI is this: Cheap and ubiquitous sensors, cameras for example, and online activity like that on social networks generate huge amounts of data. Finding the most important bits within that flood of information, and making sense of larger patterns, is a difficult problem. AI solves that problem, matching analytic capacity to the scale of the data. The result is that there is no place to hide from an authoritarian government’s gaze – that, of course, is the situation in China. 

The current challenge in the U.S. and other democracies is to leverage AI for legitimate law enforcement and national security needs, without creating tools that are too easy to abuse. The U.S. has a long history of such abuse, running from FDR and J. Edgar Hoover to the activities revealed by Edward Snowden. This time, we must act to ensure civil liberties and the rule of law ahead of such problems, rather than only in reaction to them. 

Explain why there can be no genuine AI safety without democracy and the rule of law.

I’m really pleased to see that people across society, people who have no direct involvement in the AI industry, are now talking about AI safety. It’s the right moment for that conversation, as AI systems become more capable and also increasingly permeate our activities, from commerce to education to science and beyond. 

But there is one part of the AI safety conversation that is not getting enough public attention, and it’s the one you’ve brought up: the fact that constitutional democracy is absolutely mandatory for AI safety. To see this point clearly, imagine an AI system that is safe in every way: It cannot launch cyberattacks, cannot escape human control, cannot create bioweapons, and so forth. Now picture this AI system in two settings. The first is a democracy with free and fair elections, an open marketplace of ideas, protection of civil liberties, and robust rule of law. The second is an authoritarian state, with none of those features. 

In that latter setting, our hypothetical “safe” AI is anything but safe. It will, quite reliably, be used to oppress dissent and enforce political compliance. Again, we needn’t read science fiction to understand what that looks like; it is playing out in many countries today, not only China but also Venezuela, Russia, some Gulf states and elsewhere. The take-home message is very simple: As we pursue AI safety, through technical and regulatory means, we must also work to assure that our democracies are robust and resilient as well. Real AI safety absolutely requires that.  

Assess how U.S., Chinese, and other AI companies should be held accountable for the political and economic impact of their technologies on their societies and the global community.

Here I should start by stating clearly that I am speaking in a purely personal capacity, rather than on behalf of Anthropic. That said, I think the answer to the question follows directly from the points I just made about the connection between AI safety and democracy. Any company that purports to care about AI safety – and this includes all of the major AI labs – has a responsibility to use whatever power it has to protect and enhance the health of democracy, both where the company is based and worldwide. 

It is important to acknowledge that this responsibility faces the AI industry with an extremely tricky task. Our tendency is to worry about the economic and political power of AI companies, and of course corporate power has been a recurrent problem in the U.S. and other democracies. However, at the same time, people underestimate how vulnerable corporate power – including the power of the AI labs – is to government power, especially when authoritarian tendencies seep into a democracy’s leadership. 

Over recent decades, democratic backsliding around the globe has more often involved the subjugation of corporate power by government than the opposite. And when authoritarian-leaning governments bring companies to heel, they gain access to their assets, which here include AI. So we need to watch the relationship between government and industry power very carefully and support the labs when – as we hope – they stick their necks out to defend democratic institutions and the rule of law.