Volume 2, Issue 1 - Special Issue: Profiling Elements, Risks, and Governance of Artificial Intelligence: Implications from DeepSeek; Guest Editors: Le Cheng, Tao Hong and Ming Hu

Pre-trained large language models (LLMs), epitomized by ChatGPT, have leveraged a cornucopia of “big data” to attain substantial leaps in artificial intelligence (AI). Whereas the diminishing returns from pre-training and the depletion of available training data have become evident, the post-training scaling law bolstered by “big GPU” has surfaced as an overriding strategy. Since 2024, post-trained models exemplified by o1 and DeepSeek-R1 have been widely acclaimed as successes in logic-intensive fields like advanced scientific problem-solving, serving as a bellwether for artificial general intelligence (AGI). Driven by the two cardinal elements of computing power and task-specific datasets, the data training processes of post-trained models exhibit more erratic and uncontrollable tendencies, which may be a menace to core societal domains and precipitate systemic friction vis-à-vis the existing data governance derived from pre-trained models. At this watershed moment, this article aims to conduct a comprehensive comparison of training data paradigms between pre-trained and post-trained models and to further develop cogent and favorable governance responses to mitigate emerging risks. Consequently, data security must be established as a prerequisite for AI development, and a lifecycle-based governance framework for AI training data in blended models can be introduced in the metamorphosis toward “bigger AI models”.

With the rapid development of technologies like big data, artificial intelligence (herein after AI), and blockchain, society is ushering into a new era of digital civilization. However, the same algorithms that assist in efficient decision-making for human society may also introduce issues of algorithmic discrimination. Therefore, this paper focuses on the judicial domain to deeply explore potential instances of algorithmic discrimination in AI. It identifies three key dimensions of algorithmic discrimination risks in the judicial AI domain: the ambiguity of algorithm usage boundaries, the diversity in discriminatory outcomes, and the injustice within the algorithmic environment. Building upon this, the author examines global AI governance landscapes to extract corresponding governance strategies and practical insights. Finally, a systematic regulatory approach for addressing algorithmic discrimination in judicial AI is proposed, unfolding across three levels and nine aspects: making algorithmic limitations explicit, diversifying algorithmic regulations, and justifying the algorithmic environment. This framework aims to contribute to a more reasonable, systematic, and just governance of algorithms in judicial AI.

DeepSeek marks a significant advancement in artificial intelligence (AI), ushering in a new phase of development that is reshaping the global AI landscape. AI progress has relied on the assumption that increasing computing power drives better outcomes. Yet, this focus has yielded diminishing returns, amplifying problems like market monopolies and unaddressed risks. By making its model freely available through open-sourcing, DeepSeek revitalizes the market and uses techniques like model distillation to ensure broader access to computing resources. For China, DeepSeek offers a strategic chance to close the gap with global AI leaders. However, efficiency gains in DeepSeek’s design unexpectedly increase computing demand – a phenomenon known as the Jevons paradox. Meanwhile, China grapples with limited computing resources, heightened security concerns, and stricter U.S. chip export restrictions. These difficulties underscore a deeper tension between current infrastructure and fast-evolving technology. Given computing power’s dual role, alongside China’s industrial needs and global competitive dynamics, regulation should prioritize enabling growth. This facilitative regulatory framework demands dynamic demarcation between public and private spheres in computing infrastructure governance. It also calls for advancing standardization, securing the entire supply chain, and adopting adaptable governance models. Globally, computing power governance is shifting toward “competitive cooperation,” where nations collaborate amid rivalry. China can lead by pushing for international standards, contributing its expertise to a secure, stable, and mutually beneficial global computing ecosystem.

In the digital age, the protection of privacy and personal data on social media platforms has become increasingly significant. Online behavioral data includes user interactions, fandom affiliations, and engagement records, simultaneously encompassing both defensive privacy expectations and reasonable use expectations – blurring the lines between different levels of personal data protection. The sensitivity of online behavioral data, when aggregated, generates a “digital identity” that reflects a natural person’s social identity and is closely linked to individual personality, resulting in high identifiability and privacy risks. Due to the highly commercialized use of personal data by social media platforms and the ineffectiveness of the informed consent mechanism, legal frameworks primarily focus on traditional sensitive personal data, leaving the handling of online behavioral data inadequately regulated despite its potential to cause privacy risks. By analyzing consent mechanisms and case studies from platforms such as Weibo, this research highlights how default platform settings prioritize disclosure over user control and why current informed consent mechanisms are inadequate. This study advocates for a more comprehensive boundary delineation of sensitive personal data, a renewed stratification of personal data in the digital age, and stricter informed consent mechanisms to strengthen privacy protections, aligning platform regulations with contextual privacy expectations.

Using bibliometric methods, this study systematically analyzes 6,084 AI ethics-related articles from the Web of Science Core Collection (2015–2025), capturing both recent advances and near-future directions in the field. It begins by examining publication trends, disciplinary categories, leading journals, and major contributing institutions/countries. Subsequently, co-citation (journals, authors, references) and keyword clustering methods reveal the foundational knowledge structure and highlight emerging research hotspots. The findings indicate increasing interdisciplinary convergence and international collaboration in AI ethics, with core themes focusing on algorithmic fairness, privacy and data security, ethical governance in autonomous vehicles, medical AI applications, educational technology, and challenges posed by generative AI (e.g., large language models). Burst keyword detection further shows an evolutionary shift from theoretical debates toward practical implementation strategies and regulatory framework development. Although numerous global initiatives have been introduced to guide AI ethics, broad consensus remains elusive, underscoring the need for enhanced cross-disciplinary and international cooperation. This research provides valuable insights for scholars, policymakers, and industry practitioners, laying a foundation for sustainable and responsible AI development.

Decision-makers across both technological and political fields increasingly recognize the need for AI regulation. In the context of AI governance, alignment refers to the requirement that AI systems operate in accordance with human values and interests. This article argues that misalignment is a key driver of algorithmic bias, which not only perpetuates rights infringements but also undermines AI safety, posing risks to its societal integration. This alignment imperative is rooted in the enduring principle of human dignity, a juridical concept that has evolved from its origins in Roman jurisprudence to its establishment as a cornerstone of modern constitutional democracies. Today, human dignity serves as a foundational value underpinning the rule of law. Through comparative legal analysis, this article examines how human dignity informs algorithmic governance across major jurisdictions, analyzing regulatory texts, directives, and case law addressing AI-related challenges. Despite varying implementation approaches, this paper demonstrates that human dignity can serve as a universal foundation for AI governance across cultural contexts. While the European Union prioritizes human dignity in regulating algorithmic bias, emphasizing individual rights, public interests, and human oversight, this principle extends beyond European law, offering a normative anchor for global AI governance. The article concludes with governance recommendations for the AGI era, advocating for the integration of human dignity into AI alignment. This requires both embedding dignity-preserving constraints at the technical level and developing robust assessment frameworks capable of evaluating increasingly advanced AI systems. As AI surpasses human intelligence, governance mechanisms must ensure these systems align with ethical principles, remain under meaningful human control, and operate within legally and socially acceptable boundaries.

The rapid development of Artificial Intelligence (AI) is reshaping the global landscape; however, this progress has concurrently given rise to a series of profound ethical and legal challenges. Issues such as algorithmic bias, lack of transparency, and insufficient privacy protection not only threaten technology’s credibility but also pose systemic risks to human rights protection, social stability, and even national security. Striking a dynamic balance between rapid technological innovation and effective risk regulation has emerged as a critical issue in the realm of global AI governance. In order to investigate the synergistic evolution between technological innovation and normative constraints, and to facilitate responsible development and sustainable governance of AI, this study addresses the aforementioned challenges. It integrates legal and ethical perspectives to conduct a comparative analysis alongside case studies of major global AI governance models, and to explore strategies for establishing an adaptive regulatory framework amidst the dynamic interplay of technological advancement and risk evolution. On this basis, the proposed Dynamic Interactive Double-helix Model for AI Governance–the “Technology-driven Innovation Axis” and the “Ethical-legal Constraint Axis”–facilitates the simultaneous enhancement of its security and development through dynamic interaction. It emphasizes the intrinsic driving force of technological progress as well as the embedded guarantee of ethical values, in an attempt to provide an innovative solution for the global AI governance system that combines both theoretical depth and practical feasibility.

This paper analyzes the concept of “AI for All” (AIFA), exploring its theoretical foundations, challenges, and pathways for legal implementation. AIFA encompasses three key dimensions: “AI Technology for All”, “AI Justice for All”, and “AI for All Countries”. However, the rise of generative AI – with costly subscription models – threatens these ideals, exacerbating economic and cognitive disparities. Currently, AIFA governance relies on non-binding soft laws, which lack enforceability, while hard laws remain underdeveloped. To prevent AIFA from a utopian vision descending into a dystopian reality, this paper argues for a dual approach: (1) expanding soft laws to explicitly address affordability, literacy, and international cooperation, and (2) tailoring hard laws based on national AI capacity – ranging from pricing regulations in core AI states to global cooperation for peripheral nations. By integrating soft law’s ethical guidance with hard law’s enforcement, AI governance can balance innovation with inclusivity, ensuring AI benefits all rather than deepening inequalities.