Democracy Isn’t Going To Save Us

Framing technology competition as democracy vs authoritarianism does little to help us figure out what humans require in the face of AI.

The prior essay explored the moral compromises technologists are forced to make once they realize that they exist within political power structures. There is an important downstream consequence of this I want to explore further in this essay: when AI firms align with national interests to secure competitive advantage and survival, the damage isn’t just their own moral compromise. Their alignment ends up substituting for the hard work of figuring out what values AI should actually serve.

Reading OpenAI and Anthropic’s major position papers and interviews on AI and human futures, I am struck by their utter shallowness: for documents written by some of the most capable people in the world, working on what may be the most consequential technology in human history, they say very, very little. They throw around terms like democracy, freedom, and openness. They warn (correctly) that the world could end up depending on AI systems built to serve an authoritarian state’s control over people (within its borders and beyond), and that this would be a bad thing. Their position on AI and human futures is, in a nutshell: democracy good, authoritarianism bad.

According to OpenAI’s Sam Altman,

“we face a strategic choice about what kind of world we are going to live in: Will it be one in which the United States and allied nations advance a global AI that spreads the technology’s benefits and opens access to it, or an authoritarian one, in which nations or movements that don’t share our values use AI to cement and expand their power? There is no third option — and it’s time to decide which path to take.”

OpenAI describes its OpenAI for Countries strategy as follows:

“We want to help these countries, and in the process, spread democratic AI, which means the development, use and deployment of AI that protects and incorporates long-standing democratic principles. Examples of this include the freedom for people to choose how they work with and direct AI, the prevention of government use of AI to amass control, and a free market that ensures free competition. All these things contribute to broad distribution of the benefits of AI, discourage the concentration of power, and help advance our mission. Likewise, we believe that partnering closely with the US government is the best way to advance democratic AI.”

Anthropic, for its part, issued their position paper on geopolitical AI competition, also arguing that we face a choice of two futures, one led by democracies (i.e. the United States) and the other by authoritarians (i.e. China). They state, with a tone of finality and certitude, that the world’s frontier AI systems should be shaped by “democratic values.” Sides chosen, mission accomplished. Not only have both companies aligned with the U.S. government, they have both outsourced their values framework to the tidy, crowd-pleasing slogan of “Democratic AI.”

It is no surprise that “we are building the democratic alternative to authoritarian AI” is a pitch that plays well here in Washington D.C. It supports a case for government funding, limited regulation, increased restrictions on foreign competitors, and for treating the success of their products as a matter of critical national security. As folks here in Washington like to say, D.C. doesn’t do nuance (“Senator, do you support democratic or authoritarian AI? Are you Team America or Team China?”). It is hard to really blame the tech companies for this, as taking this well-worn path was predictable—and some of what they argue is not wrong. There is a need to guard against authoritarian AI. And if these companies were only making the national security case, that would be fine. But they aren’t. They are also outsourcing their sticky ethical and moral dilemmas to democracy. And that is a big problem.

Going back to the Technologists’ Compromise, it’s not just that aligning with national political interests costs technology leaders something morally, it’s that this alignment actively substitutes for the harder work…in this case, of actually figuring out, in some rigorous and thoughtful way, what values AI ought to serve. “We are the democratic option” is not a serious answer. Rather, it serves as a convenient way of not having to answer.

Democracy as a Category Error

I don’t think the appeal to democracy is only laziness, political expedience, an attempt to secure competitive advantage, and a failure of imagination (though it is all of those things). I think it’s also a structural problem. When we organize our thinking about AI’s future around forms of government rather than around human values, we can only ever approach the issues through that lens. Unfortunately, “democratic AI” does not really tell us with any precision what we are owed by those systems increasingly making decisions about our lives, our work, our attention, our children’s education, or our access to credit, health care, or opportunity. Political frameworks simply aren’t built to answer that question, nor were they designed to.

In my work Dignity, Technology, and Global Order, I quote legal scholar Mary Ann Glendon’s writing on the nature and purpose of the American constitutional order.

“The American framers’ concept of the human person, though incomplete from a philosophical or anthropological point of view, was not inappropriate for the limited purpose of designing a federal framework within which civic life could flourish under conditions of ordered liberty. What needs to be kept in sight…is that the liberal principles enshrined in the United States’ founding documents were political principles that were never meant to serve as moral guides for all of social and private life.” (Glendon 2006)

Her point is that the framers were not doing moral philosophy. They were trying to design a federal political system that placed limits on power, protected certain individual liberties, and allowed space for a vibrant civic life to fill in the gaps. The Constitution tells us how power gets checked and how certain individual liberties are protected; it wasn’t built to tell us how to live, what we owe our neighbors, or what the good life requires.

As such, “Democratic AI” is essentially a category error. Forms of government are not designed to be the primary carriers of common human values. Asking whether an AI system emerged from a democracy or an autocracy is asking a question about the power structure it was built inside. It is not asking how the system serves the people who use it. Those need to be disaggregated. A society can be a flourishing democracy and still fail to address technology that erodes the dignity of the people living in it.

Here I want to emphasize an important point: democracy does have clear benefits over alternatives when it comes to creating the conditions under which citizens may flourish, but democratic values themselves are not very useful when it comes to telling people how to flourish.

Let’s take a closer look at democratic values. These values typically include free and fair elections, rule of law, political pluralism, checks and balances, transparency and accountability, citizen participation, responsive leadership, and individual rights such as freedom of speech, press, and assembly. Without a doubt, citizens in such a society are more likely to lead fulfilling lives than if they lived under an authoritarian regime. But when it comes to ethical questions posed by our relationship with technology, these values are not very instructive.

Here are a few examples:

• Should an AI companion or care assistant ever lie to someone for their own good, like reassuring a person with dementia that their long-dead spouse is “coming home later”?

• To what extent should people be entitled to care from another human, rather than a competent machine?

• How should we evaluate recommendation algorithms that maximize watch time and shape our attention, mood, and sense of time in ways we never meaningfully agreed to and can barely perceive?

• How should we think about platforms that amplify divisive, emotionally activating content, and do so by maximizing exactly the free expression and pluralism democratic values hold dear, even as the result is a public more fragmented, anxious, and less able to agree on basic facts?

• What does adult society owe the formative mind of a teenager scrolling short videos six hours a day and an entire industry’s business model depends on capturing attention too young to defend itself and then selling that data to advertisers?

In all of these cases, democratic values are protected. No one’s speech was suppressed, no election was stolen, no due process was denied, rule of law was followed (hey, you’re the one who clicked “I Accept” on the terms and conditions). These are not failures of democracy; these are questions democracy was never built to answer. No amount of political pluralism or checks and balances is going to tell us what constitutes and supports a flourishing life or what we owe ourselves and each other. A democratic government can protect the conditions under which good values might emerge, but it cannot substitute for the work of actually arriving at them.

Democracies Are Vulnerable

There is another weakness of democracy as it relates to this that I think is worth exploring. While democracies can provide the conditions for humans to thrive, they also provide a soft, vulnerable underbelly that make government capture by big tech arguably easier than in authoritarian systems.

This may sound counterintuitive, but I think it bears out. An authoritarian system, like that in China, concentrates power in a very tight circle. To continue to exist in an authoritarian system, companies have to become useful to the ability of people who hold that power to continue to hold that power. If they become a threat, they are quickly brought to heel. Recent history in China has shown that tech companies operate with the government’s Damoclean sword constantly above their heads. At the first signs of a tech company amassing too much power or reach, the Communist Party steps in to ensure control, both of the technology and the executives. And bottom-up influence from the tech companies is nearly impossible, especially as corruption has become more difficult. There are few doors accessible to influence politics, and the ones that do exist are heavily guarded.

A democracy, on the other hand, distributes power across many actors and institutions so that one cannot dominate the rest. But this also means there are many more doors, and they are easily accessible: lobbying, campaign finance, malleable regulatory processes, partisan press, and an active revolving door between industry and government. The openness which is one of democracy’s strengths also makes it more vulnerable to exactly the kind of capture that “democratic AI” is supposed to protect against.

These are not theoretical problems. Already the major AI tech companies are exerting significant influence on governments. The same companies holding themselves out as flag bearers of “democratic AI” are currently engaged in millions of dollars of federal and state lobbying, both directly and through Super PACs. Recent hiring flurries at these firms have prioritized government relationships as much as tech talent. When a small group of technology executives get such a heavy hand in influencing elections and deciding policy, how is that good for democracy?

A City on a Hill?

There is the added complication that the American “democratic AI” pitch isn’t landing with full authenticity in the rest of the world. The idea that the world should look to American democracy as the model for good governance assumes a shining city on a hill. It’s hard to make that pitch with a straight face at a moment when a great many Americans, let alone people watching from outside, have real doubts about the health and vitality of American democratic institutions. If the United States wants to have a central role in shaping the future of global values, our framework has to resonate with people who have good reasons to be skeptical of the Disney-fied American democracy narrative.

So where does that leave us? If forms of government can’t carry the moral weight we keep putting on them, if both the Chinese and U.S. governments and militaries aggressively pursue AI for national security advantage, if tech companies can exert such strong influence of democratic governments, and authoritarian governments exert control over AI companies, then it should be clear that the values question is not going to be answered by either governments or tech companies. Answers have to come from somewhere else, which, as far as I can figure, is civil society. And this work has to be done in the open, by people willing to ask what human dignity actually requires in the face of this technology, independent of which type of government or society is asking.

I plan to pick this up in my next essay…which I think is going to be about dignity and wine tasting.

—

Other Recommended Reading: As a new feature, whenI come across interesting work during my research, I will post it at the end. I recently read a fantastic essay by Peter Hershock on Pope Leo XIV’s Magnifica Humanitas, the subject of my first essay. He approaches this topic from the Buddhist tradition, resulting in the most thoughtful (and thought-provoking) take on the encyclical that I have read yet. You can find it here.