CHENGDU, China — Governments increasingly want to control artificial intelligence, as companies release more powerful open-source models, a new multilateral statement indicates.
The 21 APEC member economies, which include the U.S. and China, released a "Chengdu statement" on AI that emphasized respect for "security, data protection and intellectual property rights," while calling for supporting open-source development.
The statement released Thursday also specified support for models and projects "that employ strong security assurance through development and deployment."
Those details reflect how open source is moving further away from its libertarian roots, and toward one involving more state oversight. China has led recent open-source AI development, amid U.S. claims of stealing from American tech to do so.
Open-source models from Chinese companies such as DeepSeek and GLM 5.2 are free to use and download, in contrast to U.S. companies such as Anthropic that only offer closed, pay-to-use AI models.
The Chengdu statement's "explicit alignment around open-source ecosystems and compute infrastructure" confirms that Asia-Pacific is moving away from closed models to open-weight systems that are combined with state-coordinated energy, telecom and digital infrastructure, said Winston Ma, adjunct professor of law at New York University,
The APEC AI statement is the first one to include open-source cooperation at a minister level, according to Li Lecheng, China's industry and information technology minister, who chaired the leaders' meeting on Thursday.
He noted all parties recognized the risks of AI, and encouraged dialogue across the government, private sector and academia to share information about cybersecurity, supply chain resilience and online scams.
The comments come amid growing reports that Washington and Beijing seek tighter controls on homegrown AI capabilities.
"Getting 21 economies, including both China and the United States, to recognize trusted open-source AI as something worth supporting is meaningful," said Wei Sun, principal analyst, artificial intelligence, Counterpoint Research.
"The phrase 'strong security assurance' gives more security-conscious economies room to support open models while still demanding testing, transparency, data protection and deployment controls," she said. "The debate is moving beyond open vs. closed and towards: who can now build an open ecosystem that is also trusted enough for governments and enterprises to deploy."
But technology might end up moving much faster.
Janet De Silva, chair of the APEC Business Advisory Council Digital and innovation working group, earlier on Thursday urged the ministers to prepare for the quantum computing era.
"We need trusted encryption to ensure the security of our entire financial system, ensuring that quantum computing is not only a security measure but also an opportunity to bring competitiveness."
Facts Only
* The 21 APEC member economies released a "Chengdu statement" on AI.
* The statement emphasized respect for security, data protection, and intellectual property rights.
* The statement called for supporting open-source development.
* Support was specified for models and projects employing strong security assurance through development and deployment.
* Chinese companies released open-source models, such as DeepSeek and GLM 5.2.
* U.S. companies, such as Anthropic, offer closed, pay-to-use AI models.
* The statement confirmed alignment around open-source ecosystems and compute infrastructure in the Asia-Pacific.
* China's industry and information technology minister chaired the leaders' meeting.
* Parties recognized AI risks and encouraged dialogue on cybersecurity, supply chain resilience, and online scams.
Executive Summary
Full Take
The discussion surrounding open-source AI is evolving from a purely ideological debate between open and closed systems to a pragmatic negotiation over trust and governance structures. The shift toward requiring "strong security assurance" within open models suggests that the locus of control is moving from pure technological openness to externally verifiable accountability, increasingly managed by state interests. This dynamic implies that the future ecosystem will be defined not solely by code availability but by verifiable security protocols implemented across shared computational infrastructure. The focus on aligning open-source development with state-coordinated energy and digital infrastructure reveals a systemic pattern where technological dissemination is being intentionally channeled to reinforce national strategic goals regarding supply chain control and resilience. This moves the debate from "who owns the model" to "who controls the trusted deployment environment." The implication for human agency is whether this structured openness fosters genuine, decentralized innovation or merely creates a more efficient mechanism for state-mediated control over foundational technologies.
What mechanisms must be established to ensure that security assurances remain a standard for all participants rather than a variable requirement dependent on political alignment? How does the shared infrastructure model inherently mitigate the risk of creating new, centralized points of failure within the touted open ecosystem? What are the long-term consequences if trust in open systems is inextricably linked to state endorsement rather than purely technical verification?
Sentinel — Human
The text appears to be a coherent synthesis of reported statements regarding the geopolitical shift toward regulated, secure open-source AI ecosystems across the Asia-Pacific.