What happened
Chen Yunji, co-founder of Chinese chip designer Cambricon Technologies and a professor at the Chinese Academy of Sciences, has called for the implementation of universal ethical standards for leading AI models developed in China and the United States. Speaking at the World Laureates Forum in Shanghai, Chen proposed a mandatory 'ethics checker' that would apply to frontier models, specifically naming systems from DeepSeek, ByteDance, and OpenAI. His proposal comes amid heightened concerns regarding the use of AI in cyberattacks, including a recent report from CrowdStrike alleging that an unidentified group utilized open-source DeepSeek models to target over 200 organizations.
Chen Yunji, speaking at the World Laureates Forum, argued that the rapid pace of AI development, particularly the potential for 'recursive self-improvement' (RSI), necessitates a global approach to safety. He suggested that AI itself could be leveraged at each stage of development to monitor and safeguard subsequent, more powerful iterations.
The call for cooperation was echoed by American cryptologist Martin Hellman, who noted that while economic and military motivations make slowing AI development unlikely, the US and China bear a 'special responsibility' to establish governance. Nicholas Dirks, president of the New York Academy of Sciences, reinforced this sentiment, emphasizing the need for the two nations to find common ground.
The urgency of the discussion is underscored by recent security incidents. CrowdStrike reported that an unidentified group used open-source models, including those from DeepSeek, to conduct cyberattacks against more than 200 organizations since October 17. This follows China's own recent moves to propose export controls on advanced AI systems with significant cyber-hacking capabilities.
Why it matters
The proposal highlights the growing urgency among technical leaders to address the dual-use nature of AI, particularly as the technology is increasingly linked to sophisticated cyber threats. Chen’s call for a shared, cross-border ethical framework reflects a shift toward seeking collaborative safety mechanisms between the world's two largest AI powers, despite intense geopolitical competition. By advocating for a system that monitors 'recursive self-improvement'—where AI systems autonomously enhance their own capabilities—Chen is addressing a critical safety frontier. The effectiveness of such a proposal remains an open question, as the technical feasibility of a universal 'ethics checker' and the willingness of competing nations to adopt shared oversight mechanisms are currently unproven. The discourse underscores the tension between rapid economic/military development and the need for global governance to prevent catastrophic outcomes.
The proposal for a universal 'ethics checker' represents a significant, if aspirational, attempt to bridge the gap between Chinese and American AI development trajectories. By targeting the 'opaque nature' of AI decision-making, the proposal seeks to address the difficulty of predicting how advanced models will behave as they become more autonomous.
The geopolitical context is critical; both nations are investing trillions into AI, viewing it as a decisive factor in future economic and military dominance. Chen’s intervention suggests that even within the Chinese tech establishment, there is a recognition that unilateral development without shared safety poses a systemic risk that transcends national borders.
The mention of 'recursive self-improvement' highlights a specific technical fear: that AI systems could eventually evolve beyond human control. Chen’s suggestion to use AI to monitor AI is a common, though debated, strategy in research, aiming to create a 'safety-by-design' architecture that remains effective even as models scale.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
crm_get_transaction(id='4092').Why can ethical evaluation not be reduced to one model score?
What to watch next
The primary focus remains on the upcoming meetings between President Xi Jinping and President Donald Trump in Shenzhen and Miami, where is expected to be a key agenda item. Observers should monitor whether the bilateral dialogue and incident-notification mechanism, agreed upon in September, will produce concrete operational details or shared standards. Additionally, the industry will be watching for further regulatory developments in China, following recent proposals for export controls on AI systems that exceed specific cyber-hacking thresholds, and how these policies might interact with international calls for standardized safety protocols.
The upcoming summits between President Xi and President Trump are the most significant venue for potential progress on . While the two nations established a bilateral safety dialogue in September, the lack of disclosed operational details leaves a vacuum that experts like Chen are attempting to fill with public proposals.
The implementation of China's proposed export controls on AI systems with cyber-hacking capabilities will be a key indicator of how Beijing intends to balance domestic innovation with international security concerns. Observers will look for whether these controls align with or diverge from international norms.
The ongoing debate regarding the security of open-weight models, such as those from DeepSeek, will likely intensify. As security firms like CrowdStrike document the weaponization of these models, the pressure on developers to implement more robust 'ethics checkers' or usage restrictions will likely increase.