Chen Yunji, co-founder of Chinese chip designer Cambricon Technologies, proposed on Friday a universal “ethics checker” that all leading AI models would need to satisfy, the South China Morning Post reported. He named models from DeepSeek and ByteDance in China and OpenAI in the US.
Cambricon, which Chen founded in 2016 with his brother Chen Tianshi, has a market capitalisation of over 600 billion yuan (US$90 billion), the paper said. He spoke a day after CrowdStrike said a group had used an open-source DeepSeek model in attacks on more than 200 organisations.
Business Pill · SOVEREIGN AI EXPLAINED
A short explainer on sovereign AI. It teaches the general idea only and says nothing about Cambricon or this proposal.
The key insight: A chip designer’s co-founder, not a model developer, proposed on 30 October that DeepSeek, ByteDance and OpenAI models all pass one universal ethics checker, a day after CrowdStrike tied an open-source DeepSeek model to attacks on more than 200 organisations, the SCMP reported.
“All human beings and nations would need to agree on this ethics checker”, Chen said, “to prevent these large language models from doing harm to human beings in the future”. He spoke on a panel on AI safety and risks on the opening day of the World Laureates Forum in Shanghai’s Lingang Special Area.
Chen, a professor at the Chinese Academy of Sciences, warned that AI systems could soon improve themselves in a process known as recursive self-improvement. The SCMP reported he said it might be impossible to completely prevent harm, given how opaque AI decisions are.
He said he had “faith” that recursive self-improvement could be made safe by using AI itself at each stage of development to monitor the more powerful systems that follow. The report does not describe how the checker would be built, run or enforced.
The steps the South China Morning Post report places around Cambricon co-founder Chen Yunji’s 30 Oct proposal of a universal AI ethics checker. Dates and events as reported by the SCMP.
The Context Around the Panel
CrowdStrike said in a report on Thursday that an unidentified group had used an open-source DeepSeek model and another unidentified model to attack more than 200 organisations worldwide since 17 October, the SCMP reported. DeepSeek did not immediately respond to the paper.
Earlier this month Beijing proposed export controls on advanced AI systems, with extra restrictions for those that cross critical cyber-hacking thresholds, according to the report. In September the US and China agreed to set up a bilateral AI safety dialogue and an incident-notification mechanism, whose details have not been disclosed.
Turing Award recipient Martin Hellman, who spoke at the same forum, said “There is no way to stop AI in the current world, where there is too much economic and military motivation to go full speed ahead”. Nicholas Dirks, president of the New York Academy of Sciences, said the two nations have “a special responsibility”.
The Business Lever: Who Writes the Rules
The proposal comes from the co-founder of a chip designer, not from one of the model developers it names; the SCMP calls Cambricon one of China’s hottest domestic chip designers. The SCMP said the remarks come ahead of two remaining meetings this year between Xi Jinping and Donald Trump, in Shenzhen and Miami, where AI governance is expected to feature.
The Structural Read
As we read the report, the proposal names models on both sides of the US-China split and asks all nations to agree on a single checker.
As we read it, the report gives no owner, test set or enforcement route for the checker; Chen’s stated safeguard is AI monitoring the more powerful AI that follows.
As we read it, the setting matters: the SCMP places the remarks between a September bilateral AI safety dialogue and two more Xi-Trump meetings this year.
Chen Yunji, Cambricon co-founder, per the SCMP
“All human beings and nations would need to agree on this ethics checker”
Three Implications
FOR MODEL DEVELOPERS Chen named DeepSeek, ByteDance and OpenAI models as ones the checker would cover, the SCMP reported.
FOR GOVERNMENTS The SCMP reported the US and China agreed in September on an AI safety dialogue whose details have not been disclosed.
FOR SECURITY TEAMS CrowdStrike said an open-source DeepSeek model was used in attacks on more than 200 organisations since 17 October, per the SCMP.
The analysis opens its argument with: “The digital economy rests on a physical foundation that most people in that economy never see.”
As we read it, here a voice from that physical layer, a chip designer, is proposing rules for the models built on top. The framework was published in December 2025 and does not discuss Cambricon or this proposal.
What Is Not Established
The report does not say who would set or run the ethics checker, what tests it would contain, or whether any government, company or standards body has backed it. It does not say whether DeepSeek, ByteDance or OpenAI responded to the proposal.
This Week In The AI Supercycle
Execution Fault Lines, Concentrated Demand, and Geopolitical Firewalls
This week’s issue: the AI build is not short of demand for its parts. It is short of cheap money, and its end revenue rests on a narrow base of large, often venture-funded buyers.
Cambricon’s co-founder wants one ethics checker that leading Chinese and US models would all have to pass, agreed by all nations. He made the proposal a day after CrowdStrike tied an open-source DeepSeek model to attacks on more than 200 organisations.
A note on sourcing. Facts and quotations come from the South China Morning Post’s report of 31 October on the World Laureates Forum in Shanghai. The lead came from a post on X. We did not contact Cambricon, DeepSeek, ByteDance or OpenAI.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
