Politics

AI Agents' burgeoning autonomy sparks calls for radical governance reforms at China seminar

Tuesday, 2 December 2025 1:54AM UTC

A 2025 seminar at China University of Political Science and Law spotlighted the rapid rise of autonomous AI Agents with unrestricted access rights, igniting urgent debates on legal accountability, ethical boundaries, and the need for innovative international regulation.

In 2025, the rapid emergence of "AI Agents" has become the most discussed and simultaneously unsettling technological development. These AI entities are designed to significantly enhance productivity by autonomously performing tasks across digital environments. Yet, their capability to act with "unrestricted access rights" raises profound concerns about privacy, autonomy, and accountability. A pivotal seminar titled "Risks and Governance of Invasive AI: A Dialogue between Law and Technology," convened on November 28, 2025, at China University of Political Science and Law, brought together legal scholars, technical experts, policymakers, and industry representatives. The objective was to move beyond theoretical debates about AI ethics toward concretely addressing the complex challenges posed by AI Agents, particularly their expansive operational rights and governance.

The seminar was jointly hosted by the School of Civil, Commercial and Economic Law of China University of Political Science and Law, the Going Global Think Tank, and Internet Law Review. It featured three main sessions: deciphering technical risks and security mechanisms of AI Agents, defining legal and ethical dilemmas and responsibility boundaries, and exploring innovative governance models and real-world industrial practices. Participants acknowledged that while AI research is mature, governance frameworks for AI Agents remain embryonic, with urgent needs for regulatory clarity around issues such as privacy, autonomy, and the legal status of these digital actors.

From a technical perspective, experts highlighted that the foundation of AI Agent capability, unrestricted access rights, has evolved significantly. Initially intended as assistive technology for users with disabilities, these rights now enable AI Agents to autonomously interact with multiple applications, perform operations previously managed manually, and execute complex, long-duration tasks with high efficiency. However, this evolution brings two core risks: the unlimited expansion of AI rights at the system level, effectively granting full device control, and the blurring of agency, where AI Agents operate faster and more independently than human users can follow. Real-world misuse cases include automated capturing of SMS verification codes and unauthorized data migration between apps, which raise serious questions about data custody and liability under laws like China’s Cybersecurity Law.

The discussion underscored AI Agents as not mere software tools but as independent digital actors or "digital ghosts," whose activities are increasingly opaque and difficult to trace. This spurred debate on whether AI Agents should be recognized as independent subjects with distinct digital identities, separate from the natural person user's identity. Advocates argue that such recognition could foster transparency and accountability by enabling a reputation system and differentiated data pathways for Agents, rather than relying solely on user-based authorization and traditional behavioral regulation.

Legally, the seminar exposed deep uncertainties. Experts emphasized that if AI Agents' operation records cannot be traced, assigning responsibility for infringements or conflicts becomes near impossible. Divergences in industry standards further complicate matters: some associations prohibit AI Agents from using unrestricted access rights on third-party apps, while others relax controls to favour user autonomy. The principle of dual authorization, requiring consent both from users and app third parties, remains contested, reflecting potential commercial conflicts and the challenge of balancing innovation with fair competition and legal compliance.

International case law illustrates this tension. For example, in the US, the Perplexity case raised questions about whether an AI Agent acting on behalf of a user infringes platform rules and laws like the Computer Fraud and Abuse Act. The dispute reveals the trilateral conflict between users, AI Agents, and platforms, where user authorizations cannot simply substitute for platform permissions. However, AI Agents differ technologically from traditional data scraping, suggesting that existing regulations may require adaptation to new AI realities.

Beyond legal debates, the seminar considered governance innovation. Experts urged flexible, cross-domain approaches emphasizing the legality of data acquisition and the smart regulation of AI Agents. Practical issues such as cross-application data use, compliance boundaries, and technological neutrality defenses were discussed, highlighting the need for clearer legal frameworks that balance user protection with the facilitation of AI-driven services.

This seminar is a crucial milestone in AI governance dialogues, mirroring global discussions such as those held by UNESCO and the China University of Political Science and Law earlier in 2025, which focused on accelerating international consensus and scalable regulation models for AI. Similarly, recent academic work explores computational legal simulations and normative frameworks for AI governance that could help reconcile technical potential and normative constraints.

The urgency stems from AI Agents' profound systemic risks, their ability to aggregate and exploit cross-platform data, operate beyond traditional app silos, and elude conventional regulatory oversight can threaten user privacy, market order, and digital sovereignty. The unanimous call is for adaptive governance structures recognizing AI Agents as autonomous digital entities needing distinct identities and responsibilities. Achieving this requires collaborative efforts spanning law, technology, and industry, to design transparent, accountable systems that harness AI productivity while safeguarding fundamental rights and societal stability.

📌 Reference Map:

^[1] (China University of Political Science and Law) - Paragraphs 1-8, 10-14
^[2] (UNESCO) - Paragraph 9
^[3] (arXiv) - Paragraph 9
^[6] (arXiv) - Paragraph 7
^[4] (Center for the Study of Contemporary China) - Paragraph 9

Source: Noah Wire Services

More on this

https://eu.36kr.com/en/p/3577012245977475 - Please view link - unable to able to access data
https://www.unesco.org/en/articles/unesco-and-china-university-political-science-and-law-co-host-global-ai-policy-dialogue-and - In July 2025, UNESCO and China University of Political Science and Law co-hosted the 'Global AI Policy Dialogue and Collaboration Development' Forum at the Shanghai Expo Center during the World Artificial Intelligence Conference (WAIC 2025). The forum focused on frontier issues in global AI governance, facilitating exchanges across four dimensions: institutionalized dialogue, scaled application, standardized regulations, and win-win collaboration to advance international consensus. Professor Jiang Zeting of CUPL emphasized the importance of AI governance capacity-building in serving the nation's development strategy.
https://arxiv.org/abs/2510.24442 - The paper 'Law in Silico: Simulating Legal Society with LLM-Based Agents' introduces an LLM-based agent framework for simulating legal scenarios, including individual decision-making and institutional mechanisms of legislation, adjudication, and enforcement. Experiments comparing simulated crime rates with real-world data demonstrate that LLM-based agents can largely reproduce macro-level crime trends and provide insights that align with real-world observations. The study highlights the potential of AI in simulating legal systems to verify and develop legal theory and support legal administration.
https://cscc.sas.upenn.edu/events/2025/05/12/ai-governance-spillover-risks-and-global-collaboration - In May 2025, the Center for the Study of Contemporary China hosted an event titled 'AI Governance: Spillover Risks and Global Collaboration.' The panelists explored critical questions regarding AI governance, particularly concerning spillover risks that could cause global-scale consequences in the deployment of AI in various fields, including biology, US-China relations, and climate change. The discussion emphasized the need for global cooperation and preventing AI governance from becoming a tool of power rivalries.
https://arxiv.org/abs/2511.19334 - The paper 'Normative Active Inference: A Numerical Proof of Principle for a Computational and Economic Legal Analytic Approach to AI Governance' presents a computational account of how legal norms can influence the behavior of AI agents, grounded in the active inference framework informed by principles of economic legal analysis. The model aims to capture the complexity of human decision-making under legal constraints, offering a candidate mechanism for agent governance in AI systems, where agents are endowed with intentional control systems that guide real-time decisions in accordance with normative expectations.
https://arxiv.org/abs/2508.08544 - The paper 'AI Agents and the Law' examines the technical and socio-legal issues that arise as AI systems become more agentic, capable of directly executing tasks on behalf of a user. It discusses how technical conceptions of agents align with socio-legal conceptions of agency, highlighting issues such as under-specification, loyalty, and interactions with third parties. The study reveals gaps in the current computer science view of agents pertaining to legal concepts of disclosure and loyalty, and suggests a path forward for responsible AI agent development and deployment.
https://besc-conf.org/2025/program-schedule - The International Conference on Artificial Intelligence and Communication Technology (BESC 2025) scheduled a workshop titled 'The International Workshop on Artificial Intelligence for Law (WAIL 2025)' on October 16, 2025. The workshop aims to address topics such as the legal regulation of AI in international arbitration, legal characterization of reverse engineering in large-scale AI models, deconstructing compliance and exploring regulatory pathways for AI data security, and balancing data sharing and security in the era of AI, with a focus on China's smart courts.

Noah Fact Check Pro

The draft above was created using the information available at the time the story first emerged. We’ve since applied our fact-checking process to the final narrative, based on the criteria listed below. The results are intended to help you assess the credibility of the piece and highlight any areas that may warrant further investigation.

Freshness check

Score: 10

Notes: The narrative is based on a press release from the seminar held on November 28, 2025, at China University of Political Science and Law. As a press release, this typically warrants a high freshness score. No earlier versions of this content were found, indicating originality. The report includes updated data and quotes from the seminar, justifying a higher freshness score.

Quotes check

Score: 10

Notes: The direct quotes in the report are unique to this seminar and do not appear in earlier material, indicating originality.

Source reliability

Score: 9

Notes: The narrative originates from a press release by 36Kr, a reputable Chinese technology news outlet. While 36Kr is known for its coverage of technology and business news, it is not as widely recognised internationally as some other outlets.

Plausability check

Score: 10

Notes: The claims made in the report are plausible and consistent with known developments in AI governance. The seminar's focus on AI Agents and their governance aligns with ongoing global discussions on AI ethics and regulation. The technical and legal aspects discussed are coherent and reflect current concerns in the field.

Overall assessment

Verdict (FAIL, OPEN, PASS): PASS

Confidence (LOW, MEDIUM, HIGH): HIGH

Summary: The narrative is a fresh and original report based on a recent seminar, with unique quotes and data. The source is reputable, and the claims made are plausible and consistent with current discussions in AI governance. No significant credibility risks were identified.

AI governance
digital rights
technology regulation