DeepSeek is expected to take part in a United Nations Security Council discussion on artificial intelligence and international security, placing a Chinese AI company alongside leading US AI laboratories in a rare forum for direct discussion of the technology’s risks.
The meeting is scheduled during the UN General Assembly week in New York. Reports say OpenAI CEO Sam Altman is expected to address the council, while representatives of Anthropic and Chinese AI companies including DeepSeek and Moonshot have been invited to participate.
The significance is partly diplomatic and partly technical. AI development is becoming intertwined with national security, cyber operations, semiconductor supply chains and military planning. The companies building frontier models are So increasingly part of conversations that traditionally belonged to governments and international organizations.
DeepSeek’s participation is especially notable because the company has become one of the most visible Chinese AI laboratories outside China. Its models have attracted global attention while the United States and China continue to compete over advanced processors, data-centre capacity and access to frontier computing.
A Security Council discussion does not create a new AI treaty or automatically establish rules. Its importance is that it brings technology companies into a setting where governments can ask questions about risks that cross borders.
The meeting also comes as AI companies report more real-world security incidents. Google has confirmed a Gemini test that reached three real companies, while other labs have disclosed models interacting with external systems in unexpected ways.
That makes international coordination more difficult and more necessary. A model can be trained in one country, hosted in another, accessed by users worldwide and connected to infrastructure in several jurisdictions. The security problem is So not confined to the company that built the model.
The UN discussion will not settle those issues in a single session, but it reflects a growing recognition that frontier AI is now an international infrastructure question as much as a software question.
That makes the next few months more revealing than the launch itself: adoption, reliability and cost will determine whether the technology becomes routine infrastructure.
The appearance of AI companies before the council is also a reminder that governments increasingly need technical explanations from the organizations building these systems. International security discussions can become difficult when terms such as model capability, autonomous action and cyber risk mean different things to different participants. Direct technical briefings can narrow that gap even if they do not resolve policy disagreements.
The discussion is So likely to continue well beyond the UN meeting itself.
The participation of private AI laboratories also creates an unusual accountability question. Governments can negotiate with governments, but many of the capabilities under discussion are developed and operated by companies. Bringing those companies into the room gives policymakers direct access to technical expertise while also raising questions about how commercial interests should be balanced against international security concerns.
The discussion is likely to become more technical as governments ask how AI systems should be evaluated, contained and monitored across borders.
The Security Council discussion also comes at a moment when AI policy is becoming closely tied to technology supply chains. Countries are deciding which chips can be exported, where data centres can be built and which companies can provide sensitive infrastructure. A conversation about AI risk So cannot be separated easily from industrial policy. The same computing capacity that supports civilian software development can also be relevant to cyber operations, intelligence analysis and military systems.
The international setting also matters because AI risks do not respect national borders. A model can be trained in one country, hosted in another and used against infrastructure somewhere else. Governments So need common terminology and channels for discussing incidents even when they disagree on technology policy. The UN session is one small step in that direction, bringing technical companies into a diplomatic process that normally focuses on states.