Tencent’s chief AI scientist Yao Shunyu was asked at a 2026 Future Science Prize dialogue whether Tencent would keep its Hunyuan model open if it became the world’s top-ranked model, and gave no direct yes or no answer, Chinese financial outlet Yicai reported on November 8. Yicai wrote that Yao “没有直接给出肯定或否定的回答” (did not give a direct yes or no answer) and instead analysed the motives behind open-sourcing models.
Yao gave three motives for opening models, Yicai reported: winning developers when a model is not yet in the lead, sharing knowledge with the technical community, and, increasingly, safety. “The cyber abilities of [AI] models and soon [the] bio ability of models will reach a threshold where it’s actually possible to do harm”, Yao said at the symposium, the South China Morning Post reported. Yicai reported that Yao expects open and closed models to coexist.
Business Pill · API vs Open Weights
A short Business Engineer explainer on renting a model through an API versus running open weights yourself. It teaches the general idea only and does not discuss Tencent.
The key insight: Asked whether Tencent would keep Hunyuan open if it became the world’s top-ranked model, chief AI scientist Yao Shunyu gave no direct yes or no answer, Yicai reported. Yao cited three motives for opening models: attracting developers when a model is not yet in the lead, sharing knowledge with the technical community, and, increasingly, safety, and said open and closed models may coexist.
An audience member put the hypothetical to Yao, according to Yicai: if Hunyuan one day ranked first in the world, would Tencent still keep it open? Yicai introduced the question by noting that some observers say Chinese model companies open-source because they need to catch up with the leaders, a view it reported separately from Yao’s own answer.
Yao’s first motive was pragmatic, Yicai reported: for models not yet in the leading position, “开放可以吸引更多开发者使用，扩大影响力” (opening can attract more developers and expand influence). The second was the value to the technical community of sharing knowledge. Yicai wrote that a third factor is becoming more important: safety.
Yao said the team runs extensive cybersecurity tests before a release to make sure a model can be opened safely, Yicai reported, and that future models’ cyberattack and biology-related capabilities may reach a level that can cause real harm.
Yicai reported that Yao described an ideal cyber model as one with strong defensive ability, able to find vulnerabilities and help institutions improve security, but with limited ability to exploit them. Yao said attack and defence cannot be fully separated: “攻防能力并不能被彻底分离”. Even without attack content in the training data, a model may gain such abilities from better general reasoning and coding, Yicai reported.
Yao called for governance across training, release and opening, according to Yicai: careful data choices and stronger defences in training, serious automated and human safety reviews before release, and monitoring of use after launch. For future opening, Yao said the more reasonable direction may still be “开源与闭源并存” (open and closed source coexisting), Yicai reported.
Menlo Ventures reported that Chinese models grew to 89% of enterprise open-source model spend, led by Kimi (61%), DeepSeek (15%), GLM (11%) and Qwen (3%), as FourWeekMBA reported on 3 Nov 2026. Kimi plus DeepSeek = 76% is our arithmetic. Tencent’s Hunyuan is not among the model families Menlo names.
What Was Posted on X
What The Fuck: Tencent’s chief AI scientist was asked this at the 2026 Future Science Prize:
If Hunyuan becomes the WORLD’S #1 AI model, will Tencent STILL open-source it?
He did not say Yes.
Then he explained why a lab open-source its model:
>for models that are not yet in the leading position, opening them can attract more developers and expand their influence >openness can promote the development of the technical community by sharing knowledge and advancing technology >future models’ cyberattack and biology-related capabilities may become powerful enough to cause real-world harm
He says cyber defense and offense CANNOT be fully separated.
Even without explicitly training an AI to attack systems, better reasoning + coding abilities can still give it offensive cyber capabilities.
This post on X, published 8 November, drew attention to the question put to Tencent chief AI scientist Yao Shunyu about keeping a top-ranked Hunyuan open, and to the motives and the cyber and biology risks Yao described. The post says Yao did not say yes; Yicai reported that Yao gave no direct yes or no answer.
The Moat Question and an 80/20 Forecast
The dialogue also covered competition. Asked by Zhang Yaqin, a foreign academician of the Chinese Academy of Engineering, where a model company’s moat lies, Yao said AI moats sit more in the product layer than the model layer, Yicai reported, citing data, user context and trust, with WeChat as the example.
Zhang voiced support for open ecosystems and said about 80% of models may take an open route and 20% stay closed, according to Yicai: “约80%的模型可能采取开放路线，20%保持封闭”. Zhang joked that by Yao’s logic Anthropic might not be worth one or two trillion dollars, Yicai reported.
How the Remarks Surfaced
A post on X drew attention to the exchange on November 8, summarising the same motives and saying Yao “did not say Yes”. That account matches Yicai’s report. IT Home also reported on Yao’s role.
Who Yao Shunyu Is
Tencent named Yao chief AI scientist in its CEO and president’s office in December 2025, reporting to Tencent president Martin Lau, IT Home reported. In July Tencent merged its Hunyuan multimodal and large language model departments into a foundation model department run by Yao, according to IT Home. Yao joined OpenAI in 2024 and worked on its Operator and deep research agents, IT Home reported.
Where Chinese Open Models Stand
Tencent released Hy4, its new Hunyuan model, on November 4, as we reported. Menlo Ventures reported that Chinese models “grew to 89% of open-source spend, led by Kimi (61%), DeepSeek (15%), GLM (11%), and Qwen (3%)”, as we reported on November 3. Kimi and DeepSeek together account for 76% of open-source spend, by our arithmetic. Hunyuan is not among the model families Menlo names.
In the same report, open-source models handled 19% of enterprise AI tokens but took 4% of spend, Menlo said, after their token share rose from 11% to 19% between September 2025 and September 2026.
A Second Chinese Lab Voices Caveats
Four days earlier, DeepSeek research engineer Gao Xi said at the World Internet Conference in Wuzhen: “Open source may indeed amplify certain AI-related safety risks”, as we reported. Both remarks came from staff at Chinese labs whose models are released with open weights.
Update (9 Nov): Yao on closed models and Anthropic. The fuller South China Morning Post account of the same Hong Kong symposium adds Yao’s view of the closed-weights camp. According to the Post, Yao said US companies such as Claude developer Anthropic have made a “strong argument” that frontier AI models should be closed to prevent significant societal harms, while noting that open-weight models can also help bolster defences. “Maybe the best way forward is still a mix of closed and open-source [models] because you still want to empower everyone to … leverage the technology [for defence],” Yao said.
On Tencent’s own releases, Yao said: “When you release [a model], you have to be very careful about evaluation … so we collaborate with some of the best security labs in Tencent.” The Post reported that Tencent has pledged to release the weights of Hy4 “in the coming days” along with a “comprehensive system card” detailing its cyber and biological safety evaluations.
The Structural Read
As we read Yicai’s account, Yao’s first motive ties openness to position: the stated reason to open applies to models not yet in the leading position, which is the condition the audience question removed.
As we read the remarks alongside DeepSeek’s Gao Xi, staff at two Chinese labs that release open weights described safety limits on openness within the same week.
As we read Menlo Ventures’ figures, Chinese models already take 89% of enterprise open-source spend, and Hunyuan is not among the model families it names.
Yao Shunyu, Tencent chief AI scientist, as reported by the South China Morning Post
“will reach a threshold where it’s actually possible to do harm”
Three Implications
FOR TENCENT Tencent has not announced any change to how it releases Hunyuan models, and Yao said the more reasonable direction may still be open and closed source coexisting, Yicai reported.
FOR DEVELOPERS BUILDING ON CHINESE OPEN WEIGHTS Yao described pre-release cybersecurity testing and post-launch monitoring as part of governance across training, release and opening, according to Yicai.
FOR THE OPEN-SOURCE MARKET Menlo Ventures reported that open-source models handled 19% of enterprise AI tokens but took 4% of spend.
The essay says the winning pattern is to “Close the SCARCE layer (keep proprietary) + Open the ABUNDANT layer (commoditize)” and that “Timing, execution, and competitive dynamics determine whether the strategy actually works.”
As we read it, Yao’s answer is a timing question in those terms: the pragmatic motive Yicai reports applies while a model is not in the lead. The essay does not discuss Tencent or Hunyuan.
What Is Not Established
Tencent has not announced any change to how it releases Hunyuan models, and Yao did not say a top-ranked Hunyuan would be closed. We did not review a recording of the session; the account here relies on Yicai’s report, IT Home’s summary and the readable part of the South China Morning Post article. The English renderings of Yicai’s Chinese text are our translations.
This Week In The AI Supercycle
The Great Convergence of Capital, Power, and Geopolitics
This week’s issue: for two years the AI build-out was priced as a supply story, and last week as a credit story. This week it was priced as a permission story, and the answer arrived as a cost, not a ban.
Asked whether Tencent would keep Hunyuan open if it became the world’s top-ranked model, Tencent chief AI scientist Yao Shunyu gave no direct yes or no answer, Yicai reported. Yao cited three motives for opening models, from attracting developers when not in the lead to sharing knowledge and safety, and said open and closed models may coexist. Chinese models take 89% of enterprise open-source spend, Menlo Ventures reported, and Hunyuan is not among the families it names.
A note on sourcing. Yao Shunyu’s remarks are from Yicai’s 8 November 2026 report of the Future Science Prize dialogue, read in full in Chinese, and from the readable portion of the South China Morning Post’s 8 November article. Yao’s role is from IT Home’s 8 November report. English renderings of Chinese text are our translations. The Menlo Ventures figures and the DeepSeek and Hy4 details are from our earlier reporting, linked in the text. The 76% figure is our arithmetic. We learned of the remarks from a post on X. We did not contact Tencent.
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.
