As the government and industry accelerate the AI race, civil society groups continue to warn about privacy violations, labor control, the spread of surveillance technologies, and broader human rights concerns. Yet the digital justice movement has not yet grown into a broad-based social movement. While many people experience the convenience of generative AI in everyday life, they are often less aware of the risks posed by AI systems operating invisibly in areas such as hiring, welfare, finance, and public administration.
Jang Yeo-kyung, Executive Director of the Institute for Digital Rights(IDR), warns that although the harms may still appear fragmented, the rights of citizens and the principles of democracy are already being put to the test. At a time when technological optimism and narratives of national competitiveness dominate public discourse, what principles and strategies should guide the digital justice movement?
The Digital Justice Movement and Its Parallels with Climate Justice
Q. The information rights movement seems to share important similarities with the climate justice movement. As you mentioned earlier, governance problems such as the exclusion of stakeholders and the one-sided prioritization of industry interests have also been widely criticized in the making of climate policy. Unlike the climate movement, however, the fields of AI and information rights have not yet developed into a broad-based social movement capable of mobilizing civil society and the public at a larger scale.
The climate justice movement and the digital justice movement share important similarities.
First, both seek to protect social values and citizens’ lives from the impacts of advanced technologies. Second, in both fields, citizens are struggling to secure meaningful participation within systems largely shaped by powerful industries and state-led agendas. Third, both movements are increasingly concerned with the idea of a “Just Transition” as an alternative framework. Within the digital justice movement as well, growing attention is being paid to the impact of artificial intelligence on employment, leading to calls for a just transition in the age of AI.
However, compared to the climate movement, the digital justice movement has not yet matured into a broad-based popular movement. There are, in my view, two main reasons for this. One is the dominant ideology that frames AI development as a crucial driver of national economic growth and future competitiveness, reflected in narratives such as becoming one of the world’s “top 3 AI powers.” The other is the widespread influence of technological determinism and the belief in technological neutrality, the assumption that technology develops rationally on its own, and that human intervention or regulation may undermine the potential benefits of artificial intelligence.
Q. Of course, the climate agenda has also gained momentum as visible experiences of harm and crisis have accumulated, helping to accelerate the formation of a broader public sphere around the issue. Although AI-related problems are already deeply connected to everyday life, it seems that such a turning point has not yet fully arrived in the case of artificial intelligence.
Another factor is that, unlike the climate crisis, the harms caused by AI have not yet become fully visible or widely recognized in everyday life. In response to this challenge, the OECD has begun compiling case studies documenting the negative impacts and side effects of AI, and we are also collecting cases within the Korean context. From the perspective of ordinary citizens, the concrete harms of AI may still feel distant compared to the tangible sense of innovation and convenience brought by generative AI technologies.
This is also related to the different types of artificial intelligence. Broadly speaking, AI can be divided into generative AI and decision-making AI. Generative AI refers to technologies that create content, the forms of AI most people encounter in daily life. Decision-making AI, by contrast, refers to systems that evaluate, classify, or make judgments about people in various social domains. These systems are used, for example, in determining eligibility for licenses, approving bank loans, predicting crime, or evaluating students in schools.
A significant portion of decision-making AI falls into what are considered high-risk applications because they directly affect people’s rights, opportunities, and access to public goods. Yet for many citizens, the influence of these systems remains far less visible than the everyday innovations associated with generative AI. As a result, their risks often remain latent and insufficiently scrutinized.
Civil society therefore needs to continue monitoring and intervening in the operation of these high-risk, high-impact AI systems. Over time, citizens are also likely to become more aware of the implications these technologies have for human rights and public safety.
We have paid particular attention to the use of AI in hiring processes. AI recruitment tools are excluding many applicants through opaque decision-making systems. Even companies themselves often cannot explain why a candidate was rejected. For example, an applicant may be classified as having a “level 2 rating,” yet neither the employer nor the applicant understands how that rating was determined. Nevertheless, hiring decisions are made on the basis of these AI-generated assessments. We believe it is important to raise these issues together with people who have directly experienced harm or discrimination through AI-based recruitment systems.
Q. While it is important for civil society to monitor whether digital technologies are being used appropriately within social policy, the practical challenges surrounding such efforts are considerable. Are there any notable examples in which civil society intervention has successfully influenced the governance or use of digital technologies?
There have also been cases in which civil society intervention prevented potentially serious harms. One representative example concerns an AI-based immigration screening system in South Korea.
The Ministry of Justice had promoted a project to install AI systems in airport immigration areas that would recognize passengers’ faces and movements in real time and identify and track what were defined as “abnormal” or “suspicious” behaviors. We regarded the project as a serious threat to human rights and succeeded in halting it through sustained public criticism and advocacy.
However, by the time the project was suspended, subcontracted technology firms involved in the project had already trained their systems using approximately 170 million immigration records. The data included facial images, nationality information, and other forms of personal data collected from people entering and leaving the country. Because the project was structured through a competitive public procurement system, multiple subcontracted firms had already accessed and trained their AI models on these datasets. The development process itself relied on testing facial recognition error rates according to individuals’ nationalities and biometric characteristics.
In response, we pursued multiple forms of legal and institutional action. We filed a constitutional complaint, submitted petitions to the Personal Information Protection Commission, applied for personal data dispute mediation, and even requested a public-interest audit through the Board of Audit and Inspection. However, there were substantial limitations throughout the process. For example, we were denied access to personal data records needed to verify whether complainants’ information had been used, making it difficult even to establish standing as victims.
The Constitutional Court’s decision was finally issued earlier this year. The case was dismissed on the grounds that the project had already been terminated following civil society criticism and therefore no ongoing violation of fundamental rights remained. Yet the companies involved had already obtained intellectual property rights based on the project and publicly promoted the export of the technology to the United States as a successful achievement. In many ways, this illustrates how even the judiciary can become captured by forms of technological solutionism and developmentalism.
The High Threshold for Information Rights Protection
Q. Ultimately, there seem to be clear limitations to responding only after individual cases and harms have already occurred. It appears that the broader structures of state administration and the judicial system themselves make it difficult to effectively protect digital rights and information rights.
Within large-scale technological systems, it is extremely difficult for individual rights holders to raise concerns and obtain remedies. This is especially true in legal proceedings, where the burden of proof typically falls on the plaintiff to demonstrate concrete harm.
The constitutional complaint regarding the use of immigration data for AI training is a representative example. In that case, the plaintiffs were required to establish legal standing by proving that their personal data had in fact been used in the AI training process. However, the state refused to confirm whether the complainants’ data had been included in the dataset. Even within the basic structure of the judicial system, which formally relies on plaintiffs proving their own harm, it becomes extraordinarily difficult for individuals to establish violations when access to information is controlled by the state or private technology actors.
This is also why whistleblowers are regarded as critically important in digital rights and technology governance debates internationally. Many of the major controversies involving Big Tech companies in the United States and Europe became publicly known largely through the actions of internal whistleblowers.
At the same time, researchers themselves often face serious obstacles when attempting to investigate technology companies. In the case of Facebook, for example, researchers once conducted simulated advertising experiments spending money as advertisers themselves in order to examine potential biases in the platform’s advertising algorithms. In response, they received warnings that such activities could lead to legal action for interfering with business operations. Cases like this demonstrate how difficult it is for independent researchers to scrutinize the social impacts of Big Tech systems without strong legal protections and institutional support.
European AI Regulation and Korea’s Acceleration
Q. I am curious about your thoughts on the ongoing AI regulation debates in Europe. As I understand it, the EU’s AI Act has been implemented earlier and more comprehensively than comparable regulatory frameworks in Asia.
The EU’s AI Act carried significance far beyond Europe itself because it represented the world’s first comprehensive legislative framework for AI regulation. At a time when industry-led voluntary principles and self-governance mechanisms dominated global AI governance discussions, the EU’s attempt to introduce binding and enforceable regulation was highly significant. From the moment the initial draft was released, we closely examined the proposal and its implications.
Among the discussions surrounding the EU AI Act, what particularly drew our attention were the demands from civil society organizations for a human rights-based approach, as well as a series of UN documents addressing human rights principles in AI governance.
At the time, many engineers and industry actors in South Korea were deeply influenced by arguments claiming that AI could not be meaningfully regulated, or that deep learning systems were inherently impossible to explain. In response, we worked to introduce the concept of a “Human Rights-Based Approach” into domestic policy discussions. One of the initiatives we participated in together with the National Human Rights Commission of Korea was the development of human rights guidelines for artificial intelligence. We conducted research and sought to contribute actively and participatorily when the Commission formulated its policy recommendations. When the National Human Rights Commission released its AI human rights guidelines in 2022, the ongoing development of the EU AI Act also served as an important reference point.
As a result of these efforts, certain elements were partially reflected in South Korea’s Framework Act on AI. For example, the law introduced the concept of an “affected person” recognizing that individuals impacted by AI systems should have the right to receive explanations regarding decisions affecting them. It also included provisions requiring operators of high-impact AI systems to assess the potential effects of their technologies on fundamental rights. These were not included in the original draft and were later incorporated through the influence of a human rights-based approach.
However, as mentioned earlier, because policy leadership remained concentrated within industrial and technology ministries, crucial details regarding implementation were left unresolved. For instance, the law still does not clearly specify through what procedures affected individuals can meaningfully obtain explanations or challenge AI-driven decisions. In that sense, the legal framework still contains substantial gaps and limitations.
Q. At the same time, in South Korea, discussions around concepts such as an “AI Basic Society” are accelerating the introduction of AI technologies and large-scale investment at a remarkable pace. In some respects, there is even a concern that international principles and regulatory norms may struggle to keep up with the speed and realities of AI transformation in Korea. I am curious about how you view this distinctly Korean sense of acceleration and the broader state-led push toward AI-driven transformation.
In reality, the global situation is not particularly encouraging. A broader trend toward deregulation is emerging, and even existing data protection frameworks may face significant weakening. For example, discussions surrounding the GDPR (General Data Protection Regulation) increasingly emphasize reducing corporate compliance burdens and expanding data utilization, leading to calls for regulatory relaxation. The EU AI Act has also faced criticism for retreating from the level of ambition initially expected. Even provisions related to AI literacy obligations are now subject to debates over reduction or dilution within the European Union.
The problem is that deregulation in South Korea is advancing at an even faster pace than in Europe. South Korea’s Framework Act on Artificial Intelligence was already far more fragmented and limited in scope compared to the EU AI Act from the outset. Rather than comprehensively addressing AI’s impacts on human rights, labor, democracy, and broader social structures, the law was designed primarily around industrial promotion. On top of this, additional deregulation measures are now being rapidly pursued.
A representative example is the ongoing effort to revise the Personal Information Protection Act in ways that significantly expand the permissible use of personal data for AI development and deployment. In principle, personal data can only be lawfully processed with the consent of the data subject or on a clear legal basis. However, discussions are currently underway to allow personal data to be used for AI development and utilization without consent, provided such use is approved through deliberation by the Personal Information Protection Commission. Strictly speaking, this could be viewed as a dismantling of core principles underlying personal data protection law.
In fact, legislation already passed by the National Assembly allows personal data to be used without consent for the development of autonomous vehicles. Ultimately, the current trajectory suggests that foundational principles such as informational self-determination are increasingly being subordinated to industrial promotion and technological competition.
To be continued in Part 3 : AI Governance and Information Rights (Part 3)
Re-read Part 1 : AI Governace and Informtion Rights (Part 1)