On 17 July 2026, Xi Jinping delivered the opening address at the 2nd World AI Conference in Shanghai. The address touched upon the commitment to open-source use of AI, which is an alternative to the U.S led “Pax Silica” approach where AI and tech would be confined, and they can pull the plug when they choose to.
The open approach could boost technological innovation, industrial development and applications so that “all sectors and businesses can benefit from AI”, Xi said.
Xi’s address placed heavy emphasis on National security and touched on why one Country’s National Security should not override another’s, hinting at the recent Anthropic fable/Mythos debacle where Washington had restricted use of Anthropic models, citing prevention of potential “jailbreak” vulnerabilities that could enable malicious use of the models. The restriction was especially employed on foreigners accessing the AI service. Furthermore, the address conveyed that the implementation of UN based system just and equitable system” for global AI governance.
The forum saw participation from 29 countries, including Russia, Belarus, Serbia, Cuba, Brazil and Venezuela, that included 10 African and 12 Asian countries who also signed the agreement as founding members. This may well be a new entity where a Pax Silica vs. WAICO situation would arise. While the new Anthropic series of models had gained momentum in terms of identifying vulnerabilities, open-source heavy models like DeepSeek have been seen to have a steady rise; Chinese open-weight models accounted for 41% of downloads on Hugging Face this spring, surpassing U.S. models. Further evidenced by the recent Bloomberg report of DeepSeek’s Liang Topping, Anthropic’s Amodei and OpenAI’s Brockman as the Richest AI Founders.
The World Artificial Intelligence Corporation (WAICO) may well be a suitable alternative to the current U.S led bloc, which seems to be under doubt, even by the European Parliament, which summoned Anthropic to gauge the risks and reliance on AI systems developed outside of the bloc. Even more so, the AI landscape is changing vastly; every once in a while, a Chinese company releases a powerful open-weight model that is cheaper to deploy and easier to customise than its closed-end counterparts, undercutting & undermining the closed economic model of AI that U.S. firms have poured billions into.
This same month, a Beijing-based company startup, Z.ai, has climbed developer rankings with its GLM-5.2, which is said to compete with OpenAI and Anthropic models. The competition between the U.S & China is evermore, as more Chinese startups are open sourcing model’s which not only cost effective than U.S ones but are also open-source heavy. As autonomous systems are being widely adopted by China & U.S, the central brain that can automate weapons systems, China already has a central nervous system similar to Palantir. PLA’s Integrated Electronic Information System (IEIS) and the recent announcement by Anthropic in terms of their pricing model and access to Fable and Mythos, stress the open-source debate further: if anyone who has bought and paid for a service has only temporary access, what’s the point of it? The recent announcement by Anthropic in respect to the new token system also mentions that High-Tier Subscribers (Claude Max & Team Premium) retain access to Fable 5 within their existing plan but would face a 50% reduction in usage limits.
The notion of autonomous systems competing with one another is already evident, as evidenced by the recent Hugging Face breach. On July 16th, Hugging Face reported a systems breach; only this time the attacker was fully autonomous. In the blog by Hugging Face before the attacker was identified, Hugging Face made the following statement.
“The campaign was run by an autonomous agent framework (appearing to be built on an agentic security-research harness – used LLM still not known) executing many thousands of individual actions across a swarm of short-lived sandboxes, with self-migrating command-and-control staged on public services.”
The notable part of this interaction is, Hugging Face initially used commercial close ended, AI models to mitigate the attack and conduct an analysis, however, the extended use of token’s landed multiple hurdles, largely because the commercial ones guard rails couldn’t distinguish Incident responder from an attacker at which point they shifted to the open ended China made GLM 5.2, which ran on their infrastructure and was not constrained by Guard rails which made use of commercial models impossible.
OpenAI, on Tuesday 21st July 2026, released that it was, in fact, one of its models which had tried to breach Hugging Face; apparently the incident occurred during an internal evaluation which prompts models to pursue advanced exploitation using complex attack paths, in an effort to quantify their cyber capabilities. The model was able to identify the vulnerabilities of the Sandbox environment and obtain open internet access, thereby causing the incident.
Conclusion
As this article has stressed, in the new upcoming AI race, the idea of closed-ended models can only go so far; any major corporation today undoubtedly stands on the infrastructure built by Open Source; without it, all these LLM models wouldn’t even exist. However, the notion of purchasing a subscription and having it shut down without warning, seems more concerning than accidentally using poisoned LLM Models.
Open source LLM Models are precisely that, you have the whole code to edit and fix, or better yet distil from other LLM to make a new one, as U.S has alleged China of distilling its models to develop their acclaimed ones, and are hence are sanctioning those open source models.
To conclude when it comes to the question between open ended and closed-source AI models, each has its pros and cons, but for the sake of sovereignty and curbing reliance and control of foreign countries, self-hosted models seem more appropriate for the forthcoming AI Era. Furthermore, the polarisation of the Global South & West, essentially China vs USA, China as part of WAICO, will trigger collaboration between Russia and China, which may in turn end up becoming a contender to U.S led tech dominance, while Indian will have to be wary of supposed collabration between China & Pakistan in this growing age of AI, the future of open source is bright since, whatever is in the open domain, anyone can re-purpose and mold to their use case but As to how WAICO will trulyturn out, only time will tell.
