Mythos and the New Frontier of AI Cybersecurity

Mythos and the New Frontier of AI Cybersecurity

As the capabilities of cutting-edge AI models continue to advance by leaps and bounds, AI security concerns have expanded from social governance issues—such as bias, hallucinations, and misinformation—to encompass cybersecurity, the protection of critical infrastructure, and even national security. Recently, Anthropic’s Mythos model has drawn significant international attention for its enhanced capabilities in vulnerability discovery, attack chain simulation, and long-term autonomous tasks, highlighting concerns about AI-enhanced cyber capabilities. What technical barriers have these cutting-edge models overcome in terms of their cyber capabilities? How should pre-deployment security testing identify genuinely dangerous capabilities while avoiding the creation of new avenues for abuse? Do open-source models amplify risks, or do they provide the necessary conditions for independent auditing and defensive innovation? Against the backdrop of intensifying AI competition between China and the United States, can the two sides establish limited yet pragmatic cooperation? In this episode of SinoAI Insights, we will take the security controversies sparked by Mythos as a starting point to conduct an in-depth analysis of the risks, defenses, and potential for China-U.S. cooperation regarding cutting-edge AI cyber capabilities. 【Host】 ·XIAO Qian, Deputy Director, Center for International Security and Strategy, Tsinghua University 【Guests】 ·XU Wei, Vice Dean, Professor, Institute for Interdisciplinary Information Sciences, Tsinghua University ·FU Hongyu, Director, AI Governance Center and the Data Economy Center, Alibaba Research Institute 【Shownotes】 02:26 The Mythos Event as A Rehearsal for Future Frontier AI Safety Risks 05:17 Future AI Governance Focuses on "Use Control" 09:10 From Innovation to Control? How Mythos Reshapes the Landscape of US AI Policy 12:46 Chinese Perspectives on Frontier Model Evaluation, Cyber-Capability Thresholds, and AI Agent Risks 14:39 The Paradox of Model Safety: Testing May Create Even More Dangerous Models 18:05 Full-Lifecycle Governance Frameworks and China's Governance Approach 23:07 Using "Isolated Environments" to Ensure Safety and Confidentiality During Testing 25:49 Pros and Cons of the US Granting Limited Access to Mythos 27:48 Three Games Within the US Pre-Deployment Safety Review System 34:48 Risks and Value of Open-Weight Models 46:26 Prospects for US-China Coordination on Model Evaluation 【Team】 Executive Producer: XIAO Qian Producers: WANG Yexu, LIU Yuan Podcast Editor: XU Ruijia Content Assistant: CAI Leyao Graphics Designer: ZHANG Shuoning

Military AI and the Geneva Talks: Can the Future of War Still Be Governed?

As Artificial Intelligence becomes increasingly embedded in military systems, the form, tempo, and decision-making of warfare are undergoing profound transformation. From target identification and intelligence analysis to battlefield surveillance, command systems, and autonomous operations, these technological advances have also raised growing concerns over strategic stability, crisis escalation, human control, and the applicability of existing legal frameworks. Is AI simply making warfare more efficient, or quietly reshaping the logic of war itself? As military decision cycles keep shrinking and the principle of keeping humans “in the loop” faces mounting practical challenges, can humans still retain meaningful control over the use of force? At the same time, the international community has been advancing discussions on the governance of military AI within the United Nations framework. This week, diplomats and experts will reconvene in Geneva to discuss issues ranging from lethal autonomous weapons systems (LAWS) and meaningful human control to risk mitigation and emerging international norms. Yet amid intensifying great-power competition, rapid technological iteration, and growing trust deficits, what discussions are underway in Geneva? Why has forging a global consensus remained elusive? Can international law keep pace with the evolution of military AI technologies? And as human-machine collaboration becomes increasingly complex, where should the red lines of future warfare be drawn? In this episode of SinoAI Insights, we invite three distinguished scholars specializing in military AI, strategic stability, and international humanitarian law to explore how AI is reshaping battlefield dynamics, transforming the logic of war, and identifying viable pathways to global governance. 【Host】 XIAO Qian, Deputy Director, Center for International Security and Strategy, Tsinghua University 【Guests】 LI Qiang, Associate Professor, China University of Political Science and Law (CUPL); Director, Military Law Institute QI Haotian, Associate Professor, Deputy Director, Center for International Security and Peace Studies (CISAP), Peking University (PKU) JIANG Tianjiao, Associate Professor, Development Institute of Fudan University; Research Fellow, Center for Global AI Innovative Governance 【Shownotes】 02:13 How Is AI applied in the Military Domain? 05:39 How to Distinguish AI-assisted Warfare from Autonomous Warfare? 09:23 Is AI Lowering the Threshold of War? 11:53 Can Today's International Legal Framework Keep Pace with the Development of Military AI? 15:24 What Lessons Can Early Arms Control Practices Bring to the Global Governance of AI in the Military Domain? 16:55 What Are the Topics and Significance of the Geneva Talks? 19:10 What Does "Responsibility" Mean in Military AI? 21:30 Meaningful Human Control or Symbolic Human Participation? 23:44 How Do Risk Characteristics Shape AI Competition and Cooperation? 26:33 The Key to Governance Coordination: Political Will or Technological Feasibility? 30:58 How Do Stakeholders Contribute to the Global Governance of AI in the Military Domain? 35:46 Where Lie the Practical Breakthroughs for Global Governance of military AI in the Next Five Years? 【Team】 Executive Producer: XIAO Qian Producer: WANG Yexu, LIU Yuan Podcast Editor: ZHANG Shuoning Content Assistant: XU Ruijia, DU Wanhong Graphics Designer: ZHONG Junwen

The Infrastructure Behind AI: Energy, Data Centers, and the Future of Global AI Governance

As artificial intelligence becomes the core technology reshaping the global order and economic landscape, the current narrative of competition largely focuses on algorithm breakthroughs, chip computing power and application implementation. What is rarely acknowledged is that the operation of all cutting-edge AI systems relies on the support of underlying infrastructure such as data centers, power networks, cloud platforms and subsea cables. Today, AI competition has expanded from the technological dimension to an all-encompassing contest over infrastructure, energy capacity and digital sovereignty, emerging as a new frontier of great-power strategic competition. Why is energy becoming the invisible bottleneck defining the upper limit of AI development? How will the global distribution of data centers reshape the geopolitical landscape? Will the "weaponization" of infrastructure supply chains inevitably lead to the fragmentation of the global AI ecosystem? Faced with shared infrastructure security risks, can countries explore viable cooperation paths amid competition? From the unique perspective of infrastructure, this episode of SinoAI Insights delves into the power shifts, geopolitical risks and future direction of global governance in the AI era. 【Host】 XIAO Qian, Deputy Director, Center for International Security and Strategy, Tsinghua University 【Guest】 George Chen, Distinguished Visiting Professor, Schwarzman College, Tsinghua University; Senior Fellow, Centre of Civil Society and Governance, University of Hong Kong; Former Managing Director, Public Policy, Greater China at Meta 【Shownotes】 02:11 AI competition is fundamentally rooted in infrastructure competition 06:11 Key factors influencing AI infrastructure investment site selection 10:00 Fragmentation risks of AI development and digital colonialism concerns 13:06 Strategic status of data centers and three models of digital sovereignty 17:21 Geopolitical risks to AI infrastructure investment: The Gulf case 20:56 Competition landscape of data centers in Asia-Pacific 28:43 Appeal of China's AI technologies to the Global South countries 31:43 Cooperation space and risk control paths in China-U.S. AI competition 35:20 From platform era to AI era 41:37 Attitude differences towards AI between Chinese and U.S. Gen Z and their impacts 【Team】 Executive Producer: XIAO Qian Producer: WANG Yexu, LIU Yuan Podcast Editor: DU Wanhong Content Assistant: ZHONG Junwen Graphics Designer: CAI Leyao

Can AI Dialogue Rebuild Trust? China-U.S. Cooperation After Resuming Intergovernmental Dialogue on AI

Over the past two years, discussions about artificial intelligence between China and the United States have increasingly been shaped by the narrative of strategic competition. Yet at the same time, AI is also creating shared risks that neither country can manage alone. During U.S President Donald Trump's visit to China, both leaders reached a consensus to resume governmental dialogue on AI. Can AI dialogue provide a new anchor for stabilizing China-U.S. relations? In an environment marked by intense competition, expanding controls, and insufficient mutual trust, what role can Track II exchanges play? As Artificial Intelligence is rapidly becoming a defining technology of the 21st century, maintaining channels of communication may itself become an important form of risk mitigation. In the second episode of SinoAI Insight, scholars and industry experts with extensive experience in Track II dialogues explore the significance of restarting China-U.S. intergovernmental AI dialogue, the practical impact of Track II dialogues, technological competition and the future of the industrial ecosystems, as well as prospects for bilateral cooperation on AI. 【Host】 ·XIAO Qian, Deputy Director, Center for International Security and Strategy, Tsinghua University 【Guests】 ·LU Chuanying, Professor, Vice Dean, School of Political Science and International Relations, Tongji University; Vice President, Shanghai Association for Artificial Intelligence and Social Development ·YAO Xu, Secretary-General, Center for Global AI Innovative Governance; Associate Professor, Fudan Development Institute, Fudan University ·FU Hongyu, Director, AI Governance Center and the Data Economy Center, Alibaba Research Institute 【Shownotes】 01:55 The significance of restarting China-U.S. intergovernmental dialogue on AI 05:37 The risks of embedding AI into early warning and C2 systems 08:42 Dialogue can increase policy predictability amid competition 10:30 Cooperation between tech communities: frontier model safety and open-source AI governance 13:05 Is Track II dialogue merely talking? 19:32 The importance of building shared terminology 21:31 How industry can contribute first-hand experience to Track II dialogue 27:25 AI dialogue is showing signs of institutionalization 29:21 Prospects for China-U.S. AI cooperation over the next five years 33:10 The signal released by restarting dialogue: competition and cooperation are not antagonistic against each other 40:40 Most important principle that should guide China-US relations in the AI era 【Team】 Executive Producer: XIAO Qian Producer: LIU Yuan, WANG Yexu Podcast Editor: CAI Leyao, LI Kaijuan Content Assistant: ZHANG Shuoning Graphics Designer: DU Wanhong

Between Rivalry and Risks: AI and the Future of U.S.-China Relations

As artificial intelligence rapidly evolves into the defining technology of the 21st century, it is reshaping the landscape of U.S.-China relations in unprecedented ways. How do leading Chinese experts assess the intersection of AI competition and shared security risks? What realistic opportunities exist for bilateral cooperation amid escalating strategic rivalry? And what frameworks can manage the existential threats posed by unregulated AI development? Recorded against the backdrop of President Donald Trump's visit to China, the first episode of SinoAI Insights, a new podcast launched by the Center for International Security and Strategy, Tsinghua University (CISS), explores the dual nature of AI as both a driver of great-power competition and a catalyst for necessary cooperation, offering a comprehensive Chinese perspective on the future of U.S.-China relations in the AI era. 【Host】 XIAO Qian, Deputy Director, CISS, Tsinghua University 【Guests】 SUN Chenghao, Fellow, CISS, Tsinghua University JIANG Tianjiao, Associate Professor, Development Institute; Research Fellow, Center for Global AI Innovative Governance, Fudan University GU Dengchen, Research expert on China's industrial policy; Non-resident Fellow, CISS, Tsinghua University 【Shownotes】 【01:53】 Expectations for the AI agenda at the summit 【05:30】 When AI meets the "nuclear button" 【07:23】 Shared U.S.-China AI risks under the shadow of chip export controls 【09:54】 Beyond "human-in-the-loop" paradigm 【12:55】 How to overcome the security dilemma of defensive R&D 【14:28】 Non-state actors and the misuse of AI 【17:53】 The three most urgent AI security frontiers shared by China and the U.S. 【20:49】 Chinese industry perspectives on AI risk and governance 【24:15】 Distillation: Technical Tool or Political ladder? 【27:17】 Will technological competition undermine scientific collaboration and open innovation? 【29:55】 In favor of functional cooperation to a unified governance architecture 【34:00】 What kind of crisis management mechanisms are more resilient? 【36:48】 Are Chinese scholars still optimistic about the future of AI governance? 【Team】 Executive Producer: XIAO Qian Producer: WANG Yexu, LIU Yuan Podcast Editor: WANG Yexu Content Assistant: ZHONG Junwen Graphics Designer: XU Ruijia