DDemocracy is colliding with the technology of artificial intelligence. Judging by the reaction of the audience at the recent World Forum on Democracy in Strasbourg, the general expectation is that things will get even worse for democracy. There is another story. While there are certainly risks to democracy from AI, there are also opportunities.
We published a book called Rewiring Democracy: How AI will Transform Politics, Government, and Citizenship.. In it, we provide a clear perspective on how AI is undermining trust in our information ecosystem, how biased AI use can harm members of a democracy, and how elected officials with authoritarian tendencies can use AI to consolidate their power. But it also provides a positive example of how AI is transforming democratic governance and politics for the better.
Here are four such stories currently unfolding around the world. It shows how AI is being used by some to make democracy better, stronger and more people-sensitive.
Japan
Last year, then 33-year-old engineer Takahiro Yasuno was a leading candidate in the Tokyo gubernatorial election. Running as an independent candidate, he ultimately finished fifth out of a field of 56 candidates, largely due to his unprecedented use of a certified AI avatar. That avatar answered 8,600 questions from voters in a 17-day continuous YouTube livestream, capturing the attention of campaign innovators around the world.
Two months ago, Anno was elected to the Japanese Senate, once again leveraging the power of AI to engage voters, this time answering more than 20,000 questions. His new party, Team Mirai, is also an AI-powered civic technology shop that develops software aimed at improving governance and making it more participatory. The party is using some of Japan’s public funding for political parties to build the “Future Congress” app, which allows voters to express their opinions and ask questions about parliamentary bills, and use AI to organize their wording. The party has pledged that its members will direct questions in committee hearings based on public input.
Brazil
Brazil is notoriously litigious and has even more lawyers per capita than the United States. Courts are chronically overwhelmed with litigation, and the resulting backlog costs the government billions of dollars. It is estimated that Brazil’s federal government spends about 1.6% of its annual GDP on court operations, and an additional 2.5% to 3% of its GDP on court-ordered payments from cases it loses.
Since at least 2019, the Brazilian government has been actively deploying AI to automate procedures across the judiciary. Rather than making judicial decisions, AI helps distribute caseloads, conduct legal research, transcribe hearings, identify duplicate filings, prepare initial signature orders, cluster similar cases for joint review, and more, all to make the justice system work more efficiently. And the results are important. For example, Brazil’s Federal Supreme Court backlog fell to its lowest level in 33 years in 2025.
While it is clear that courts are recognizing the efficiency benefits of leveraging AI, there is an addendum to court AI implementation projects over the past five years. That means litigators are also using these tools. Lawyers are filing cases in Brazilian courts at an unprecedented rate with the help of AI, with the number of new cases increasing by nearly 40% over the past five years.
For Brazilian litigants, regaining the upper hand in this arms race is not necessarily a bad thing. It has been argued that litigation, particularly litigation against governments, is an important form of citizen participation that is essential to the self-governing functioning of democracies. Court systems in other democracies should learn from Brazil’s experience and strive to leverage technology to maximize court bandwidth and liquidity in processing cases.
Germany
Now we’re moving to Europe to innovate in voter information. Since 2002, the German Federal Office for Civic Education has been running a nonpartisan voting guide called Wahl-o-Mat. Officials will convene an editorial team of 24 young voters (those under the age of 26 and selected for diversity) and experts in science and education to create a set of 80 questions. The question will be submitted to all registered German political parties. Answers will be narrowed down to 38 key topics and published online in quiz format. Voters can use it to identify the political party whose platform they most identify with.
For the past two years, an external group has been developing an AI-powered alternative to the official Wahl-o-Mat guide. First up was Wahlweise, a product of German AI company AIUI. Second, students at the Technical University of Munich introduced a conversational AI system called Wahl.chat. Over 150,000 people used this tool within the first four months. In both cases, instead of having to read static web pages about the positions of various political parties, citizens can more easily obtain the same information tailored to their personal interests and questions by interacting with an AI system.
But German researchers studying the reliability of such AI tools ahead of Germany’s 2025 federal elections have raised major concerns about bias and “hallucinations,” or the ability of AI tools to fabricate false information. The researchers recognized the potential of this technology to increase voter information and party transparency, and recommended adopting scientific evaluations comparable to those used in the official tools of the Civic Education Agency to improve and institutionalize this technology.
US
Finally, the United States, specifically California, home of CalMatters, a nonprofit, nonpartisan news organization. Since 2023, the company’s Digital Democracy Project has collected all the public statements of California elected officials, including all floor speeches, comments on committees and social media posts, voting records, legislation, and campaign contributions, and made all that information available on a free online platform.
This year, CalMatters released new features that take this type of citizen surveillance capability even further forward. Its AI Tip Sheets feature uses AI to search through all of this data, looking for anomalies such as changes in voting rankings associated with large campaign donations. These anomalies appear on web pages that journalists can access, providing story ideas and sources of data and analysis to drive further reporting.
This does not mean that AI will replace human journalists. It’s a citizen watchdog organization that uses technology to provide evidence-based insights to human reporters. And it’s no coincidence that this innovation came from a new kind of media institution: nonprofit news agencies. As the decline of newspaper business models continues to diminish the oversight capabilities of the Fourth Estate, this type of technical assistance is a valuable contribution that helps a small number of human journalists retain some of their reach and influence the nation’s democracy that depends on them.
After newsletter promotion
TThese are just four of the many examples around the world of AI helping to strengthen democracy. What they all have in common is that technology decentralizes power rather than centralizing it. In all four cases, they are being used to support, rather than replace, those carrying out democratic tasks: politics in Japan, litigation in Brazil, voting in Germany, watchdog journalism in California.
In none of these cases is the AI doing anything that humans cannot do fully competently. However, in all of these cases, there is not enough human resources to do the work themselves. Fully trusted AI can bridge the gap, amplifying the power of public servants and citizens, improving efficiency, and fostering government and public engagement.
One barrier to realizing this vision more broadly is the AI market itself. The core technology is primarily developed and sold by large US technology companies. We do not know the details of their development: what materials they were trained on, what guardrails are designed to shape their behavior, what biases and values are encoded into their systems. And even worse, we have no say in the choices associated with those details or how they change over time. In many cases, it is an unacceptable risk to use these commercially-proprietary AI systems in a democratic context.
To address this, we have been advocating for the development of “public AI” for many years. These are models and AI systems that are developed under democratic control and deployed in the public interest, rather than sold by companies for the benefit of shareholders. This movement is growing around the world.
Switzerland recently released the world’s most powerful and fully realized public AI model. It’s called Apertus and was developed jointly by the Swiss government and ETH Zurich. The government has made it completely open source, open data, open code, open weight, and free for anyone to use. No illegally obtained copyrighted material was used in the training. We do not exploit low-wage human workers from the Global South. Its performance is roughly on par with the leading companies from a year ago, and is more than sufficient for many applications. And it shows that we don’t need to spend trillions of dollars to create these models. Apertus has taken a major step toward realizing its vision of an alternative to enterprise AI managed by leading technology.
AI technology is not without cost and risk. We’re not here to minimize them. However, this technology also has significant advantages.
AI is inherently empowering, allowing the humans behind it to scale up what they want to do. It can strengthen authoritarianism just as easily as it can strengthen democracy. It’s up to us to steer technology for the better. Society will be better off if more citizen watchdogs and litigators use AI to strengthen their powers to oversee and hold governments to account, when more political parties and election officials use AI to meaningfully engage and inform voters, and when more governments provide democratic alternatives to big tech’s AI services.
What is giving me hope right now?
Nathan: I look forward to seeing examples of people using technology to fight for what they believe in. In addition to the above, we see political activists in the United States using AI to make door-to-door campaigns more effective, municipal engineers in Brazil using data to create oversight of local governments, and countless other examples around the world.
Bruce: There is real potential for AI technology to decentralize power. Efforts are currently being concentrated because building these frontier models is extremely costly. However, their costs have fallen dramatically, increasing the need for smaller, more agile models. One of the most important things we can do is break down technology monopolies, and a more decentralized AI ecosystem could help us do that.
-
Nathan E. Sanders is a data scientist at Harvard University’s Berkman Klein Center and co-author with Bruce Schneier of Rewiring Democracy: How AI Will Transform Our Politics, Government, and Citizenship. Bruce Schneier is a security engineer who teaches at the Harvard Kennedy School at Harvard University.
