1. Introduction
In the months leading up to the US election often the question was raised how AI tools might influence and modify the ways people experience the presidential election. The release of the sophisticated chatbot ChatGPT 4o in May 2024 raised questions about how this tool might influence the path to the US presidential elections later that year. As far as we know by now, some experiments with AI-generated audio and video content were to be seen during the last stages of the election campaigns. But when it comes to organizing elections there are almost no traces of AI usage in this otherwise highly computerized and data-intensive process.Footnote 1 Nevertheless, modern elections can be conceived as a socio-technical system – the electoral process in many ways relies on technological solutions: voter information, identification and registration, collecting, verifying and counting the votes are processes that are in many ways supported by technology. In some countries, these steps are conducted by using innovative hard- and software embedded in a set of administrative routines to increase the efficiency of election management. But how do those devices and processes become part of the official legislation and can finally be deployed during this sensitive and important democratic procedure? As the development of new technologies in this sector is not easy to predict and the ability of public administration is to a certain degree limited in terms of resources and cost, election officials need to figure out ways to identify, test and decide for (or against) the use of new technologies. In the US (and elsewhere, too) this is a decentralized process, happening on state and municipal level. Thus many different strategies can be observed, but many of them deal with the question of integrating technologies that evolve over time and open up new possibilities for organizing the electoral process. Currently, digital technologies for casting, verifying and counting the vote on location have replaced older forms of electronic voting. These now mostly retired electronic voting machines had replaced mechanical voting machines – often as a direct consequence of the infamous “Florida Recount” during the 2000 presidential elections which led to the “Help America Vote Act” (HAVA) of 2002. The funds provided by this legislative initiative can be seen as a watershed moment for a nationwide modernization of the electoral process on different administrative levels (Election Assistance Commission, 2023). Recent developments include the usage of portable devices to integrate the routines in the polling places, the development of machines to provide voter-verifiable ballots as well as the implementation of automated ballot counting and reviewing processes.
With public interest (and anxiety) growing when it comes to the use of AI systems, the perspective of the recent history of electoral technology and engineering seems to be a perspective on past developments – yet it is helpful to investigate further in these processes to also focus on things that still lie ahead. The main idea of this article is the notion that there are already some experiences with regulatory design able to structure of future development and implementation of technology into the electoral process. And even though these rules and regulations do not aim directly towards AI tools and services, the legislative measures in question might deploy a framework how to regulate AI systems in the future. A key element of this framework is the institutional structure of electoral management as a social form of controlling technological developments. At this point it is helpful to review some additions to the research literature on electoral management bodies (EMBs). Scholars often try to assess the general performance of these institutions with regard to their effects for the quality of political systems as a whole. Early publications like Lopez-Pintor (2000) discuss the histo-rical development and spread of EMBs, and identify several types of EMBs ranging from permanent, independent authorities to temporary, in part state-run organizations. In general, EMBs have become a regular feature in modern democracies and help to increase legitimacy and trust in electoral processes (James, Reference James2020; Langford et. al. Reference Langford, Schiel and Wilson2021; Mozaffar & Schedler, Reference Mozaffar and Schedler2002). Regarding the capacity of EMBs, the focus is “on “their” ability to perform functions and achieve their goals” (Garnett, Reference Garnett2019, p.336), usually meaning the ability to support the democratic functions of elections itself.
Large comparative studies on EMBs may draw a distinct picture of the importance of this type of organization for the support and perceived quality of democracy, the role of election technology often is no major subject in those surveys. While Brown, Hale and King (Reference Brown, Hale and King2019) cover a broad range of practical aspects running EMBs, only very few contributions address technological aspects like the “overseas vote” (Kalisa, Reference Kalisa and Brown2019), the procurement of election technology (Bennett, Reference Bennett and Brown2019), or electronic voter registration (Becker, Reference Becker and Brown2019). Until now, the ability to monitor, evaluate or even improve the electoral process is rarely considered as an integral goal or function of EMBs. Many authors state substantial differences between the institutional arrangements and are convinced that EMBs play an important role for the stability of political systems. The usage of technology as a means for organizing the electoral process seems is not regarded a relevant issue within this body of literature.
The relevant case study, the State of California, has developed a robust system for integrating innovative technology into the electoral process and is also able to change and modernize its rules and regulations. The technologies currently used are more static, hardware-based and usually do not include algorithmic systems, but the overall structure of the process can also function as a blueprint for regulating more dynamic algorithm-based or even AI-based technologies. The following article tries to explore the experiences with organizing and modernizing the electoral process as something close to a much needed form of experimental legal regulation that can also be applied on AI tools, systems and platforms. The underlying case is reviewed from a political science point of view and thus does not deal with fine-grained legal developments but with the institutional structure that obviously is able to guarantee legitimacy of and trust in the electoral process while being transformed by technology. These factors are also necessary when it comes to developing and regulating AI systems. Based on a review of literature, the paper describes the main structures of electoral management in countries that have successfully modernized elements of the electoral process. The condensed case study of California is using means derived from the method of “process tracing,” outlining the origins and relations of various institutions of electoral management on state level. After describing the Californian case as a “regulatory ecosystem,” the paper explores and discusses the possibilities of using these measures as a blueprint for conceiving regulatory structures for AI systems.
Before diving more thoroughly into the subject, it might be helpful to think about how political actors usually react to the challenges of regulating AI systems. How can political governance of AI systems actually work? When reviewing the early practices of taming an AI, a kind of “folk theory” of political regulation of algorithms can be observed. Political actors (like a parliament, a federal or state government or an executive branch) seem to assume AI systems as kind of a personal embodiment whom they want to behave to their service. But of course, this does not apply.
In a typical form of calling for help, political actors then bring into place a team of experts, a commission, or a public-private initiative which shall find solutions to control AI systems, following the demands of the political actor. These often semi-institutionalized assistants routinely develop a set of recommendations with basic principles, rules and regulations that political actors might use for creating and passing a law.Footnote 2 In this multi-step-process the typical time-lag of the political process immediately becomes clear – already the establishment of supplemental actors for law-making bodies take a lot of time, and then the activity of an expert board is only beginning. The production of practical knowledge about a new, previously unknown object of regulation is also time-intensive and still it is not marking the end of this overly simplified policy learning process. After producing (and sometimes publishing) a report on new possible forms of regulating, the technology in question may have reached its next stage of development. And still – a report is not a law. To adopt these expertly developed recommendations as a framework for legal jurisdiction, the legislative process sets the rules for drafting and finalizing the actual content of a legal measure. Under the circumstances, it is hard to imagine how a functional legislative process with AI systems as an object of law-making might actually work. The long and winding road to the EU AI Act only gives a first impression of how this constellation may look like in practice.Footnote 3
To contrast the uncertain future of AI-oriented law-making, this article shifts the attention to the legal efforts that already have been pushed forward by the State of California to incrementally regulate voting technologies of different sorts. In the following I will argue, that the institutional framework and the set of rules conceived by Californian lawmakers can be regarded as a blueprint for crea-ting a functional structure that can be applied to AI systems as well. Before presenting key elements of regulatory actions on voting technology, I will discuss the complicated nature of modifying and modernizing electoral processes in various countries. While Germany still is focusing on old-school, analog modes of casting the vote, legal activities concentrate only on granular specifications to the electoral process. Other countries like Brazil or Estonia invest heavily in the modernization of voting by simultaneously inventing new institutional setting that help to create stability and trust in the electoral system.
2. Elections: Just another algorithm?
One of the most important democratic processes itself constantly is subject to transformation – political elections as an instrument for selecting representatives in parliaments or for directly fil-ling offices follow similar patterns, but the specific implementation changes over time. This dynamic affects electoral law – in Germany, for example, it recently led to the extinction of the infamous Überhangmandate, regularly discusses the determination of the voting age or questions specific methods of voting.Footnote 4 But it also effects different elements of its practical implementation – such as the design and production of ballots, the organization of absentee ballots or the creation of additional opportunities for voting beyond election day and away from polling stations. Until the ruling of the Federal Constitutional Court in 2008, the use of “electronic voting machines” was also a starting point for change; after problems in dealing with voting computers, such high requirements were formulated that this option of election modernization is de facto excluded in Germany (cf. Bieber, Reference Bieber, Roessing and Podschuweit2013, Reference Bieber, Schünemann and Kneuer2019).Footnote 5
Despite the focus on analog forms of voting, recent discussions about modifying electoral law referred to the analogy of elections being some kind of algorithm. In a contribution for the German online periodical Verfassungsblog, Maximilian Steinbeis argued that especially in Germany elections follow the model of an algorithm (Wettstreit und Algorithmus, 1.10.2021; https://verfassungsblog.de/wettstreit-und-algorithmus/).
It is helpful to emphasize on the use of the algorithm as a metaphor. Not only finding a result and transforming the ballots cast by the voters into the composition of the parliament can be conceived as an algorithm. Also the dynamic interplay between electoral law, the institutional background of the electoral system with its relevant actors constitute the broader configuration of the electoral system as a whole. In Germany, the Federal Returning Officer (Bundeswahlleiter:in) is simultaneously the president of the Federal Statistical Office (Statistisches Bundesamt). Another important commission is the Federal Election Caucus (Bundeswahlausschuss), a small group of members of parliament and judges, who oversee the organizational process leading up to the election. These two main actors only are constituted immediately before the formal electoral process is started, usually about half a year before election day. The more practical tasks of preparing the ballots, recruiting staff for election day, and organizing the counting process is done on local level, where municipalities are responsible for organizing the actual voting process. Thus, in Germany there is no standing commission, authority or court responsible for election administration. This huge task is fulfilled only in a restricted period shortly before and after election day. Members of the Federal Election Caucus are usually replaced every four years, and the position of the Federal Returning Officer is dependent on the role of leading the Federal Statistical Office. Although the actual people in charge may change, the institutional context of election organization in Germany remains stable, with a continuous interplay of the relevant actors guaranteed. But, under the circumstances, substantial innovations of the voting processes have proved to be very unlikely.
2.1 The rise of electronic voting solutions
On the other hand, the introduction of digital elections (in Estonia) or the gradual technologization of the electoral process (in Brazil or India) reveal that it is apparently possible to develop a functioning, legally secure and at the same time socially accepted procedure for conducting political elections using technical elements. When it comes to single case studies on specific political systems and their conduct of the electoral process, technology may well shift into the center of attention. This is especially true regarding the innovative technologies implemented in the electoral process of Estonia, displaying cutting-edge solutions for digital, internet-based voting. The report of the Office for Democratic Institutions and Human Rights (ODIHR) Election Expert Team provides an overview on the state of internet voting efforts during the last national parliamentary elections (ODIHR, 2023). Interestingly enough, the body of literature on the development and usage of Estonian digital voting system is heavily focused on the technical aspects of the process. It seems that institutional innovations – like creating a strong set of EMBs – do not find the attention of international scholars excited about the technology in use (and its weaknesses and failures). Since the first digital election in 2005, a broad set of studies has been published, describing closely the different stages of development. While Alvarez, Hall and Trechsel (Reference Alvarez, Hall and Trechsel2009) portray the basic functions of the voting system and they also emphasize the role of the institutional background – but not so much with regard to Estonia, but as a reason why other states fail in creating and implementing innovative technologies: Often “a lack of cohesive election laws that facilitate online transactions” is a massive roadblock when it comes to modernizing the electoral process. With every successful iteration of the Estonian internet voting, the range of studies is extended. Vassil, Solvak, Vinkel, Trechsel and Alvarez (Reference Vassil, Solvak, Vinkel, Trechsel and Alvarez2016) describe usage patterns during the first decade of digital voting, claiming that internet voting can be an “inclusive technology,” but needs time to spread among the electorate. Ehin et. al. (Reference Ehin, Solvak, Willemson and Vinkel2022) cover an even longer period of internet voting (2005–2019), adding a lot of data on user demographics and the broad acceptance in the electorate over time. Also, they discuss the role of trust in the process of internet voting for the likelihood to actually cast a vote online: (…) In sum, trust in i-voting is the key factor explaining usage even after the wide diffusion of this voting mode in the population. The most recent study takes into account the latest modifications to Estonian internet voting system “IVXV” and examines closely current technological features, reported security breaches and also improvements made in the process of reviewing new versions of the technology in use (Arafat, Reference Arafat2025). In general, the institutional framework is considered to be capable of organizing internet-based elections and generating trust in the system. Nevertheless, ongoing technological development may threaten the integrity of the process, especially the future possibilities of quantum computing are likely to be used to compromise electronic citizen IDs used as an organizational backbone of internet voting systems.
Besides Estonia, only a few other countries are subject of systematic, and continuous academic research on digitally supported election systems.Footnote 6 While Estonia is a small country with only about 1 Million eligible voters, Brazil (156 Million) and India (970 Million) are examples at the other end of the spectrum. Brazil is an interesting case for two reasons: on the one hand, low-tech electronic voting machines are used to collect the vote (Bieber, Reference Bieber, Schünemann and Kneuer2019), on the other hand the judiciary branch is fully responsible for the electoral management, it oversees and develops the electoral process (Câmara, de Castro & Oliveira, Reference Câmara, de Castro and Oliveira2022). Other than Estonia, Brazil does not collect votes via Internet but uses kiosk-style devices to record the voter´s choice locally. To increase transparency in the act of voting, since 2021 paper ballots are printed additionally, they can be checked by the voters and are stored for auditing (Tribunal Superior Eleitoral, Reference Tribunal Superior Eleitoral2025). Similar to Brazil, India relies on a low-tech approach when it comes to electronic voting machines (EVMs). The devices are used only in polling stations, the votes cast are stored locally and the machines are sealed after use, before the votes are being counted in a separate routine (Herstatt & Herstatt Reference Sadhu, Ghosh, Nandi, Majumder, Zaman and Ghosh2014: p.10–15, Nazary, Reference Nazary2021: p.2929–2930). For future elections, India plans to extend the use of voting technology including Internet voting solutions for the expatriate community. In both countries systems are subject to critique because of possible security breaches or intransparency regarding the hard- and software in use. Nevertheless, the institutions controlling the electoral process have been successful to moderate controversies and were able to introduce improvements or new routines for testing the systems (Sadhu, Ghosh & Nandi, Reference Sadhu, Ghosh, Nandi, Majumder, Zaman and Ghosh2025). Again, the need for a combination of technological and political structures for organizing the electoral process becomes clear.
The rise and development of new encryption technologies also fuels the discussion about the modernization of elections, simultaneously adding some impulses from Computer Science to the otherwise largely social/political science-driven debate. Especially advances in blockchain-based systems help modeling a more secure voter registration and a less vulnerable electoral process (Arafat, Reference Arafat2025; Jafar & Ab Aziz, Reference Jafar and Ab Aziz2020; Sadhu et al., Reference Sadhu, Ghosh, Nandi, Majumder, Zaman and Ghosh2025). Praise for the opportunities of distributed blockchain-based approaches is mixed with critique regarding the relative immatureness of these systems, its ecological footprint, or the decentral structure of blockchain-technology, which is difficult to monitor and govern (Park, Specter, Narula & Rivest, Reference Park, Specter, Narula and Rivest2021). The most recent studies also take into account the so-called “quantum threat,” when a new generation of hyper-accelerated quantum computers might be able to break any current cryptography. Under these conditions, theoretically no security can be guaranteed for digital voting solutions as they rely heavily on encryption (Arafat, Reference Arafat2025). Nevertheless, there still is a lot of speculation and insecurity how the “quantum threat” will play out in practice and whether proposed responses like “Post-Quantum cryptography” (PQC) are needed to counter these threats. An interesting fact to conclude this review of current – technical – research literature is that systems based on Artificial Intelligence like Large Language Models, conversational agents, or other forms of agentic AI are not subject of the discussion on the future of digital voting solutions.
2.2 Estonia, Brazil, California: Installing a powerful electoral management
In the following, I will not focus on the specific technological innovations, but on the rules and regulations political actors implement to sustain control over the process – my thesis is that the design of repeatable, rule-guided routines based on technical expertise can also help to ensure effective political regulation of AI systems for decision-making or other purposes relevant to society.
Even if there are major differences between the election technologies used in each case, the institutional arrangements reveal similar structures – a multi-layered system in which technology and legal development are intertwined is implemented by public actors in cooperation with private companies and secures the processes of election organization. In Brazil, for example, the Supreme Electoral Court (Tribunal Superior Eleitoral, tse.jus.br) shields the entire electoral process at the highest state level; in Estonia, a state electoral authority (Riigi valimisteenistus) has been established with special tasks for the organization and development of elections; and in the USA, the Election Assistance Commission (eac.gov) supports executive authorities that are responsible for supervising elections in the states. In all these cases, professional associations and private companies come in and are tasked with various testing operations. Also, the legal frameworks evolve over time – when technological innovations are integrated into the electoral process or particular difficulties arise.
State supervisory bodies usually provide the central regulatory framework for the conduct of poli-tical elections, while the actual implementation is then carried out by local actors. In the run-up to the election period, these electoral management bodies conduct various testing and inspection procedures to certify the voting equipment and technologies used and approve their use at the polling station. In California, which plays a pioneering role among the US states, a Voting Modernization Board was set up in 2002 as a result of the Help America Vote Act, which essentially oversees the distribution of funding for the further development of voting infrastructure in individual counties. The Board is under the purview of the Secretary of State, and numerous state administrative processes converge in this election office. At the same time, the incumbent assumes the role of Chief Elections Officer and is responsible for other subordinate authority structures. Since 2005, the Office of Voting Systems Technology Assessment (OVSTA, https://www.sos.ca.gov/elections/ovsta) has been more operationally involved in the development process and is primarily responsible for two tasks: the investigation, testing and certification of voting systems as well as the approval of ballot printing machines and the production and distribution of ballot papers in the election process.
In Brazil, the various supervisory and development tasks at federal level are concentrated in the Supreme Electoral Court. There, too, the various tasks are spread across different bodies; most recently, a Commission for Transparency in Electoral Processes (Comissão de Transparência das Eleições) was set up in 2021 to guarantee the security of electoral processes in conjunction with a supervisory authority (Observatório de Transparência Eleitoral). As in California, the voting technologies used here also undergo an elaborate and carefully documented process with various stages of testing. This includes the use of IT experts appointed by the parties, who can review and comment on the software code a year before the election. On election day itself, an “integrity test” takes place in which the voting equipment used is checked in a public place by an independent test facility.
The measures in California, Brazil and Estonia prove that there are already viable testing systems in place able to deal with complex technological systems and their development over time. In this way, political elections can clearly be accompanied, monitored and further developed. A by-product of these institutionally bound regulatory systems is the increase in trust in a formerly “black box,” which can also strengthen the stability of the political system. The challenges facing policymakers are very similar to the situation in the field of Artificial Intelligence: how can a fluid, evolving system that relies on large amounts of data and continuously integrates new technologies be controlled? Obviously, the integration of a complex technological setting into political processes can only succeed if a self-contained system of actors and procedures is created that is capable of organizing a permanent “process evaluation” and involving numerous experts from different fields. At the same time, there needs to be a strong link to political structures, ideally at a high judicial level in the case of political elections. Political regulation of AI does not necessarily have to be tied to supreme court structures; instead, an institutional anchoring of the hard- and software in use seems to be more appropriate.
If this task is mastered successfully, then trust in new, complicated technologies that are not immediately understandable for citizens might be created over time – a by-product that should not be underestimated for the stabilization of democratic communities and a major challenge for the actions of political actors in the digital transformation.
3. The case of California: Creating a regulatory ecosystem
As noted before, California is among the US states with the most elaborate electoral system manage-ment structure – thus it is a helpful example for regulating innovative technology. Oversight and control of the electoral process in California is vested in the office of the Secretary of State, Shirley N. Weber.Footnote 7
The Secretary of State, pursuant to the authority granted in Elections Code Section 19232 withdraws certification and conditional approval of all California voting systems, in whole or in part, not tested and certified to the California Voting System Standards (CVSS), effective August 27, 2019. Elections Code section 19232 states that ‘the Secretary of State may decertify any voting system or part thereof, determined to be defective, obsolete, or unacceptable.’ (https://www.sos.ca.gov/elections/ovsta/certification-and-approval/voting-systems-approved-use-california)
The invention of new technologies is a vital element of the state´s industry not only since the rise of computing – from the early 20th century on the film industry, the development of energy systems, traffic and infrastructure, or aviation and aerospace technology set the stage for modern digital innovations connected to “Silicon Valley,” like transistors, microchips, computer languages, hard- and software development, data storage, the Internet, social media and Artificial Intelligence.
Against this backdrop it is not very surprising that California takes an outstanding role in regard to usage and regulation of voting technology. Over the last two decades, California has developed a distributed legal infrastructure that is capable of tracking, monitoring and modernizing the electoral system by consistently improving voting technology and its political regulation.Footnote 8 The starting point dates back to 2002, when the Voting Modernization Board (VMB) was implemented, a direct consequence from the federal Help America Vote Act (Voting Modernization Board, Reference Voting Modernization Board2002). The VMB is active until today, its basic task is distributing money to counties throughout the state in order to fund the modernization of voting equipment. The money stems initially from the federal Help America Vote Act, which initially provided $ 195 Million to the state of California. Since then, additional state funding was available for upgrading and developing voting technology on county level.
The funding process stimulated increased activity of election officials and led to the implementation of new administrative units. In 2005 the Office of Voting Systems Technology Assessment (OVSTA) was created, it is responsible for testing and certifying voting equipment of all sorts. According to the OVSTA-Website, there are two main tasks:
1) voting system examination, testing, and certification for use in California elections; and 2) overseeing the approval of ballot printers, as well as authorizing and monitoring the manufacture and distribution of ballots for an election. Over the years, that scope has expanded to voting technologies including ePollBooks and Remote Accessible Vote by Mail Systems. (https://www.sos.ca.gov/elections/ovsta)
The constant inflow of money fueled the widespread use of voting technology by counties awarded with money from successful VMB applications. In the early phase of the funding process, many of the innovative systems were so-called “Direct Recording Electronic” (DRE) voting machines, storing the votes internally, often without paper-based records. In several studies, these devices revealed vulnerabilities that sometimes lead to the de-certification even of newly implemented systems (Blaze, Reference Blaze2020). Over time, OVSTA developed a complex and differentiated framework for testing, certifying, and reviewing voting technology (see Figure 1). Within this framework also routines were established to cooperate with voting technology vendors, escrow facilities (for equipment testing), election observers, and pollworkers. The assessment process is an ongoing procedure, which has helped to organize and structure the modernization of voting technology in California. Problems with voting technology still occur – especially when even more steps in the electoral process are supported by new hard- and software solutions. Thus, OVSTA recently has expanded its scope and now handles also ePollBooks (organizing routines at the polling place and controlling the voting register) and Remote Accessible Vote by Mail systems.
Figure 1. Review and testing for certification process overview, Office of Voting Systems Technology Assessment (OVSTA, https://www.sos.ca.gov/elections/ovsta).
Nevertheless, OVSTA is an important actor within the ecosystem of electoral organization in California and definitely has helped to establish oversight in a very diverse process of technology development.
With the notion of a lack of transparency and security breaches in voting technology, but also regarding the growing external threats to election security, the Secretary of State in 2018 installed an “Office of Election Cybersecurity” to tackle problems specifically connected to digital security issues, including dis- and misinformation, and even “electoral rumor control” (Freed, Reference Freed2018).Footnote 9 Long before creating this specialized unit, imperfect voting technology had become a target for watchdog organizations. The California Voter Foundation is a very early example for “civic tech activism,” monitoring digital voting information and technology since the 1990s (www.calvoter.org). Since then, the widespread use of different means of voting technology has led to a more diverse field of watchdog organizations that help to inform the electorate about the ongoing process of modernizing the vote. These actors contribute to building a more robust public discourse about elections and their technological components. The most advanced pool of information about voting technology is maintained by VerifiedVoting.org, another non-profit-organization dedicated to these issues in the electoral process. The main tool of the group is the so-called “Verifier,” a powerful database with information of election day equipment in use in all of the United States. Their data shows a slow transition to paper-based or paper-including voting procedures, leaving behind the more tech-oriented approaches of the 1990s and 2000s.
To sum up, the Help America Vote Act of 2002 functioned as a starting point for the development of a small administrative ecosystem dedicated to support the distribution of modern voting technology throughout the State of California. The Secretary of State as key actor in the executive branch of government guarantees democratic legitimacy, while several units take care for the execution of specialized tasks such as money distribution and funding (Voting Modernization Board), technology certification and control (Office of Voting Systems Technology Assessment), and security measures (Election Security Task Force). Additionally, these units engage and cooperate with commercial actors (escrow facilities) and members of the civil society (watchdog organizations) to extent their outreach into the field of operation.
A closer look into the routines of the certification process established by OVSTA reveals a certain level of agility and flexibility in technology assessment. As voting technology continues to evolve over time and its implementation might stretch even across election cycles, political actors had to establish processes that keep track of technical advancements and are able to repeat and iterate the certification process. The result is a standardized routine that can be adapted to various forms of technology, depending which part of the electoral process is under review (i.e. voting systems for polling, election day-tools (PollBooks), vote-by-mail-solutions).
Thereby constituted is a mostly linear process of regulation, but with “pockets” of testing, feedback and refinement by participants (phases 2 and 3). Especially the rules for the testing process (phase 2) foster an in-depth cooperation between vendors and certification authorities, in part supported by additional actors with technical expertise.Footnote 10 To some extent, bringing in the public for hearing and commenting on the test results (phase 4) resembles the inclusion of stakeholders in the process.Footnote 11
The influx of new solutions to the field of voting technology obviously affected the certification process on the administrative end. Electoral law-making now reveals patterns similar to elements of a regulatory sandbox, Footnote 12 offering developers and vendors of voting technology a publicly controlled environment before bringing a new product into the market. Although there is no standing definition of the sandbox-concept, there is a rough consensus about its building blocks:
Regulatory sandboxes are experimental legal regimes which waive, modify national regulatory requirements (or implementation) or provide bespoke guidance on a temporary basis and for a limited number of actors in order to support businesses in their innovation endeavors. A regulatory sandbox offers safe testbeds for innovative products and services without putting the whole system at risk. Sandboxing aims to promote thus the advancement of technology, new policy solutions through the promotion of collaborative regulation, and novel compliance initiatives between innovators and regulators. (Ranchordas, Reference Ranchordas2021)
The workflow of the certification process shows some similarities to the sandbox approach outlined above. “Sandboxing” has proven to be efficient especially when it comes to the integration of new technological solutions (Ruschemeier, Reference Ruschemeier2023). Conceiving California´s regulatory efforts as a long-term activity one recognizes the evolution of a productive network of actors that has its origin in the Help America Vote Act from 2002. This legislative reaction to the flawed presidential election of 2002 directed money to the federal states and helped to constitute committees like the Voting Modernization Board (Reference Voting Modernization Board2002), and later on the Office of Voting Systems Technology Assessment (OVSTA) as a key player in the process described here. Overseen by the Secretary of State, a continuous development process led to passing the Californian Voting System Standards (CVSS). First issued in 2014, it still provides the key framework for “specifications and requirements against which voting systems shall be tested” (CVSS 2014: 2). Since 2018, the Office of Election Cybersecurity infuses more technical expertise to the system, as do non-profit watchdog organizations California Voter Foundation (calvoter.org) and VerifiedVoting (verifiedvoting.org) as stakeholders representing the so-called “civic tech”-sector. In the final stages of certification, public hearings reach out to the broader public so citizens as the targeted “end user” of a voting system can also have an impression of (and share their opinion about) technical solutions they will cast their vote with in future elections.
Since more than two decades California has grown a regulatory ecosystem that integrates a set of different stakeholders in a process that has created pockets for developing and testing new voting technologies. The eventful history of machine- and computer-assisted voting in the USA might be considered as a blueprint for applying a similar framework to other technologies, which itself change over time and thus require a flexible configuration of political oversight and leadership, effective administrative structures, technological expertise, and a connection to public discourse and scrutiny.
4. Conclusion: Learning from a regulatory ecosystem
Currently, there is plenty of regulative activity under way in California, addressing various fields like transparency, privacy, safety, and even election integrity (Anderson, Reem & Taddayon, Reference Anderson, Reem and Taddayon2024). Yet it is not clear, which government agencies are involved on state level, where there is an active exchange with federal actors and whether there are specialized structures for organizing, monitoring and controlling these processes.Footnote 13
By taking into account the last two and a half decades of building a regulatory ecosystem for the certification and implementation of voting technology, there might be some lessons to learn for lawmakers dealing with the affordances of AI systems regulation. To conclude this article, I will try to collect a few starting points.
- The electoral process itself experiences a “digital transformation” and thus new forms of testing and certifying voting technology had to be developed. Up to now, those technologies are more static, hardware-based and usually do not include algorithmic systems. But this may change over time, and then there might be an overlap between voting technology assessment and the certification of AI systems.
- The basic idea of “governing” emerging, unfinished technologies brings new and sometimes innovative actors in the political realm – and reveals difficulties to implement these new structures into the political process. Once these effort are successful, a new regulatory ecosystem is constituted, thereby changing and re-organizing institutional structures.
- This process takes time and the final design of such an ecosystem cannot be projected in the beginning stages. Actors, routines and rules emerge and not necessarily stay in exact the same shape over time They might also be organized cross-sectorally to combine the experiences and competencies of several branches of government.
- Robust institutional structures are a keystone for regulating innovations, also financial resources and personnel is needed to continuously monitor the process. A continuous funding is necessary to uphold the regulating bodies. Especially when it comes to long-term transformations, different sources of funding (federal level, state/local level) have to be combined.
- Not everything is about speeding up the legal process. On the contrary, creating patterns, pockets and loops that allow or enforce iterations in using and testing new technologies, can be a helpful strategy to include different kinds of observatory actors. The back-and-forth between public committees, commercial testing facilities and also non-government organizations has proven to stabilize the usage of new hard- and software-solutions in important democratic processes.
- Without firm political leadership it will be difficult or even impossible to implement innovative regulations for emerging and quickly evolving technologies. Developing trust for all stakeholders involved is a crucial task that can be achieved by functioning processes and routines, but a clear “line of command” is helpful to initially establish a functional framework. While for voting technology the Secretary of State´s Office seems to be the natural choice for this role, other digital technologies might by overseen by different government agencies (in California, i.e. Department of Technology, Office of Systems Integration, Office of Data and Innovation). As soon as relevant political processes are affected by using new technologies, leading branches of government (Governor´s Office, Secretary of State´s Office) should be integrated directly into regulatory structures and procedures.
Acknowledgements
This work was supported by the Villa Aurora & Thomas Mann House e.V. for funding a Thomas-Mann-Fellowship in Pacific Palisades which allowed extensive field work in California. The author would like to thank the reviewers for their thoughtful comments that helped to sharpen the rationale of the article. Furthermore, the author thanks Leonie Klammer who supported the development of the article and the process of revising it.
Competing interests
The author declares no competing interests.
Christoph Bieber is Professor for Political Science at the NRW School of Governance, University of Duisburg-Essen, Germany. His main area of research is the political system of Germany and its digital transformation. Since October 2021 he is delegated to the Center for Advanced Internet Studies (CAIS.nrw) in Bochum, where he is heading the Research Program “Digital Democratic Innovations.” Key research projects include the political aspects of Smart Cities and forms of digital decision making. Bieber has published widely on the effects of digitalization for political actors and institutions. His books include “Nach Obama” (with Klaus Kamps, 2017), “Politik digital. Online zum Waehler” (2010) and “Unter Piraten. Erkundungen in einer neuen politischen Arena” (ed. with Claus Leggewie, 2012). He blogs at http://internetundpolitik.wordpress.com. In 2022, he received a Thomas-Mann-Fellowship for expanding his research on ethics and the smart city in Los Angeles.
Open access
