Alibaba's Joe Tsai makes the open-source AI case at Wuzhen as Beijing eyes UN rules

WUZHEN, China: Alibaba Group chairman Joe Tsai opened the World Internet Conference Wuzhen Summit on Tuesday. He argued that open-source models are the only way to build an AI industry that lasts. He spoke in Zhejiang province at an event backed by China's Cyberspace Administration and the provincial government, the South China Morning Post reported.

"Only through open source can AI become popularised, and only through popularisation can we create a sustainable, inclusive intelligent era," Tsai said, according to SCMP. The report said he pointed to growing convergence on open ecosystems among domestic tech firms. AI demand is climbing while computing power stays scarce.

That is one thread of a wider push. On 27 October, SCMP reported that Beijing was making the case at a global forum for UN-led governance of artificial intelligence. The Wuzhen meeting opens the week after that pitch. The two fit together: China wants to be seen setting the rules for how models get shared, not just how they get built.

The commercial backdrop matters. Alibaba is doubling down on open-weight releases. Tsai's remark lands while US labs are raising record sums and taking positions in the regulatory fight. Palantir reported third-quarter results on Monday that beat estimates. Revenue topped $2 billion for the first time and net income more than doubled to $1.13 billion. US commercial revenue jumped 148% to $984 million, ahead of US government revenue of $953 million, according to CNBC. CEO Alex Karp told CNBC he has supported open-weight models and has called for nationalising AI labs.

Gulf money keeps flowing in parallel. OpenAI signed agreements covering energy, government and investment with Abu Dhabi entities, CNBC reported on Monday. These include a deal with ADNOC and G42 to deploy its frontier technology across the oil company's operations. The Abu Dhabi Media Office confirmed the arrangements. OpenAI is seeking up to $20 billion in a new round co-led by UAE investors, CNBC reported.

Governance work is also moving below the summit level.

A paper posted to arXiv on 29 October documented a pilot expert workshop on "violet teaming." This is a method for testing biodefense-relevant AI capabilities across development, misuse and implementation stages. The authors, Steph Batalis and four co-authors, said the exercise surfaced non-obvious misuse pathways. It also expanded possible interventions beyond technical safeguards. It is a small study, but it shows the practical testing question that UN-level talks will eventually have to answer.

Who writes the rules

The Wuzhen stage is not the only venue. DPA reported on 31 October that the EU and Canada forged an "alliance for the future" at a Montreal summit. That is a signal that middle powers want their own channels alongside UN processes. The Diplomat, meanwhile, published an analysis on 31 October of how Kazakhstan's economic agenda under President Tokayev shapes its international partnerships. Neither piece settles the AI question, but both show governments shopping for forums.

Security researchers, for their part, keep documenting what happens when governance lags. The Hacker News reported on 2 November that Palo Alto Networks Unit 42 had tracked a campaign, CL-UNK-1211, against Linux systems at government and state-affiliated bodies between October 2025 and April 2026. The campaign used PHP and shell scripts that appeared to be generated with large language models or automated agents. The same outlet reported a day earlier on NetamLoader, malware aimed at government Windows 10 machines in Asia.

None of this is on the Wuzhen agenda. But it is the argument Tsai's audience will hear from the other side of the table: open models spread capability, and capability spreads to attackers too. The summit runs through the week. Whether it produces anything more than statements is the open question.