1. What can we learn from computer systems for democracy?

I think of democracy as an information system. It’s a way to convert all of our individual preferences into group outcomes, in a way that’s fair and equitable. It’s how we determine our policies, and who makes policy in our name: what our tax rates are, what our foreign policy looks like, what we deem legal and illegal.

And — this is interesting — it’s an archaic information system. The modern constitutional democracy was invented in the mid-eighteenth century, using mid-eighteenth-century technology. Because travel and communications were so hard, we had to pick one of us—a representative–to go all the way across the country and pass laws in our name. Because the cost of coordination was so expensive, we built a system that uses competition — elections — to figure out what to do. A lot of these assumptions no longer hold true in the second quarter of the twenty-first century.

There’s a lot we can learn from modern information systems — IT systems — that we can apply to the information system that is democracy.

2. Where are democracy’s biggest vulnerabilities today? How is AI changing them?

Generally, I think the biggest vulnerability is in the rigidity of the rules of democracy, and how effective different parties are at exploiting them. This leads to a rigidity of democracy itself, because the powerful people who benefit from the current set of rules don’t want to change them. Which makes any change, except for minor changes around the edges, impossible.

These can be explicit rules, such as the makeup of a legislature. But they can also be implicit rules that are a result of implicit rules. I’m thinking of the US two-party system, which is an effect of our first-past-the-post voting system and the entrenched power of the two existing parties. Or money in US politics, which is the result of a bunch of structural artifacts of the US system.

This is a huge vulnerability, as we as a species are facing enormous problems that span countries: climate change and the massive migrations it will cause, for one. But we can’t solve that problem unless we can reverse the collapse of democracies around the world, and we can’t solve that because of the rules we’re bound to.

AI has the potential to upend all of this, but I have no idea how. Because AI is the first technology to replace human cognition, as opposed to human power, it’ll change every aspect of the information system that is democracy. I just published a book, Rewiring Democracy, that discusses how the technology will change elections, legislation, government administration, the courts, and citizenship. There are lots of examples in the book about AI changing democracy, both positively and negatively, from all around the world. But I can’t predict how it will all shake out. I try to be optimistic.

3. You define hacking as exploiting the rules of a system in unintended ways. What are the most important ways AI could be used to „hack“ democratic processes and are we prepared to defend against that?

No, we’re not. I think of hacking as finding loopholes in the rules. Think of a tax loophole: It’s something that the rules permit, but are unanticipated and unwanted by the designers. One of the problems with democracy right now is that the rich and powerful have hacked it so thoroughly — have found and exploited so many legal and regulatory loopholes — that the system is failing. I even wrote a book about this: “A Hacker’s Mind”.

AI can supercharge this process. The technology is really good at finding loopholes in sets of rules. Mostly computer code right now, but the tax code is just another set of algorithms to analyze and exploit. I worry that AI-empowered people, already rich and powerful people, will be able to find and exploit a wholly different level of loopholes, and that they will further erode the systems of democracy — and our trust in those systems.

4. Democracies are often slow to change by design. Is that slowness now a liability, especially in a world where AI-powered hacks can rapidly exploit systemic weaknesses?

It always was a liability, but it mattered less before than it does today. Today, technology changes so fast, and technology companies are so nimble, that they can run rings around government processes. Democracies need to be more agile, creating policies at the speed of tech so they’re not always playing catch-up. Easier said than done, because — as you pointed out — that the slowness is part of the design.

The one place that democracies are nimble is in the regulatory agencies, and that’s probably the near-term solution. Legislators need to pass laws giving these agencies broad authority to quickly implement detailed rules as circumstances change.

5. What can be done against the erosion of trust in democracies, and what role can AI play in this?

As I mentioned earlier, I just published a new book: “Rewiring Democracy. How AI Will Transform Our Politics, Government, and Citizenshop.” t’s more than deep fakes; it’s a book about democracy much more broadly. And what my co-author and I found is that AI is doing positive things in democracies around the world: in legislatures, in bureaucracies, in the courts. AI can be a force for democracy, and for trust in democracy, if used properly.

The way to think of AI is as a power enhancement. If the people using AI want more democracy, and more trust in democracy, AI can help. If the people using AI want less democracy, AI can help with that as well. We have examples of AI being used to make government agencies more responsible, to reduce dependencies on paid lobbyists, to make courts fairer, and to give people more of a voice in their political process. All of this increases trust in democracy.

6. Do you think AI could subtly shift democratic systems toward authoritarianism , even without explicit intent – particularly through increased surveillance, control, or centralization of power?

Of course. It could do that slowly and subtly, or it could do that quickly and explicitly. We are seeing democracies backsliding toward authoritarianism broadly across the globe, and AI will play a part in that. One aspect is the technology’s capabilities in surveillance systems, or — even more ominously — as eavesdropping spies. Another is the ability to centralize control. We saw that in the US, as Elon Musk’s DOGE used AI to efficiently cancel government programs — something that it would have taken much longer for people to do. Bureaucracies move slowly; AIs that take the place of thousands of bureaucrats can change instantly. This is something we talk about in our book: AI that centralizes power hurts democracy, whereas AI that democratizes power helps democracy.

7. What scenario at the intersection of AI and political, economic, or military power worries you the most? And which one, perhaps surprisingly, gives you the most hope?

Right now, I am most worried about the consolidation of AI in the hands of just a few corporate monopolies, mostly in the US and mostly run by white male billionaires. This monopolization is incredibly dangerous to democracy and to society more broadly, and I want antitrust enforcement, and regulation in general, to break this power center. And here I am looking at the EU to lead, because nothing will happen in the US along these lines anytime soon.

On the hope side, AI has some pretty amazing capabilities as a moderator, a mediator, and a consensus builder. There are experiments in many countries along these lines, like the 420,000-person online citizen assembly that MAKE.org facilitated in Germany in 2024. AI has the potential to help people become more engaged in politics and policy, which means a better democracy for everyone.

8. What do you think democracy might look like in the future (say 10–20 years from now)? Do you think it will endure in the long run at all?

Reimagining democracy for this century is something I think about a lot. It’s pretty clear to me that the government systems developed in the mid-eighteenth century are ill-suited to the mid-twenty-first century. But while I have a lot of ideas of how to redesign democracy from scratch, I am much less confident about how we can get from here to there. That’s the realm of politics, not tech.

And while I am not optimistic about the corporations that control technology or even the current popularity of anti-democratic politicians, at least in the near term, I am confident about people and about democracy. Over the centuries, society has become more fair, more equitable, more democratic. So while we might have a bumpy couple of decades ahead of us, I think democracy will endure.

9. How is AI changing the cybersecurity playing field. That is, how is it shifting the balance between attackers and defense?

((laughs)) That’s a big question, and an important one. There’s really no quick answer; it’s shifting so many aspects of both cyberattack and cyberdefense that it’s hard to predict the new balance. I recently wrote an essay on this exact question, which I urge people who want a more detailed answer.

10. If AI becomes highly effective at finding vulnerabilities in complex systems, does that change our basic assumptions about what “secure” even means?

No, of course not. “Secure” is fundamentally a human term that’s way older than technology. The basic assumptions about what secure, or safe, means is psychological, not technological. Technology can change how likely it is that we feel secure, of course, but it can’t change the meaning of the word.

And your question assumed that AIs finding vulnerabilities in complex systems makes us less secure. What if AIs found all the vulnerabilities in those systems and fixed them? Then it would make us more secure. This is the arms race between attacker and defender, and it’s not obvious how AI will change it long-term.

11. As AI becomes embedded in more everyday systems we rely on, from infrastructure to communication, what are the key risks to public trust if those systems fail or are manipulated?

12. Quantum computing is often described as a potential game changer for cryptography and security. From your perspective, how serious is the threat it poses to current security systems and are we moving fast enough to prepare for a post-quantum world?

This is another massive question whose answer would take an entire essay. Fortunately, I’ve written it before, and I urge readers to click on the link to find out more. Briefly, there is a lot of physics and engineering between today and a useful quantum computer, and no one has any idea if and when it will be feasible to build one. I like to say that the engineering problems are hard. And by “hard,” we don’t know if it’s “land a person on the surface of the moon – hard,” or “land a person on the surface of the sun – hard.” Both are hard, but they’re very different levels of difficulty.

At the same time, we have made great progress in designing quantum-safe cryptographic algorithms. So, the math is way ahead of the physics. What we need to do is to migrate to these new algorithms and, more generally, build crypto agility into our systems. Google just announced that they are migrating by xxxx. This is the right approach; do it now, when there’s no reason to panic.