# DeepSeek invited to make a statement at UN Security Council AI meeting > Source: > Published: 2026-09-22 10:33:00+00:00 # DeepSeek invited to make a statement at UN Security Council AI meeting **Reuters says DeepSeek and other Chinese AI developers were invited to make statements at the September 23rd meeting, but DeepSeek's attendance, delegate and remarks remain unconfirmed.** By [RuntimeWire Staff](https://runtimewire.com/author/runtimewire-staff) ยท Published Primary source: [Reuters](https://www.reuters.com/world/asia-pacific/deepseek-brief-un-security-council-ai-this-week-sources-say-2026-09-22/) ## Why it matters DeepSeek's invitation gives a Chinese AI lab that distributes model weights and sells managed API access a possible place in a global security discussion that includes closed-model US rivals. DeepSeek's attendance, delegate and position on safety or open-weight governance remain unconfirmed. [DeepSeek](https://deepseek.com/?ref=runtimewire), the Hangzhou AI developer founded by Liang Wenfeng, was invited to make a statement at a United Nations Security Council meeting on artificial intelligence and international security scheduled for September 23rd, [Reuters reported](https://www.reuters.com/world/asia-pacific/deepseek-brief-un-security-council-ai-this-week-sources-say-2026-09-22/?ref=runtimewire). Other Chinese AI developers, including Moonshot AI, also received invitations. DeepSeek's invitation did not amount to confirmed attendance. As of September 22nd, Reuters had not identified DeepSeek's delegate or confirmed that DeepSeek would deliver a statement or attend in person. Plans remained fluid one day before the meeting. DeepSeek was invited to make a statement at the September 23rd Security Council meeting on AI and international security, according to Reuters. The report did not identify what DeepSeek would say or what council members wanted to hear from DeepSeek. Reuters separately reported that OpenAI CEO Sam Altman planned to brief the council in person. Diplomats expected a high-level Anthropic presence, [according to Reuters](https://www.marketscreener.com/news/openai-s-sam-altman-to-brief-un-security-council-next-week-during--ce785adadf88f52c?ref=runtimewire), although Anthropic's participation remained unconfirmed. The [Security Council](https://main.un.org/securitycouncil/en?ref=runtimewire) has 15 members and primary responsibility for maintaining international peace and security. France, which holds the council presidency in September, convened the meeting. A French concept note reviewed by Reuters said the session would examine risks involving loss of control and the use of advanced AI models in actions affecting international peace and security. ### Liang built DeepSeek after High-Flyer Liang founded DeepSeek in 2023 after building High-Flyer, an AI-focused quantitative investment manager. The [Associated Press](https://apnews.com/article/deepseek-founder-liang-wenfeng-china-ai-0673d5c39d90108189cc31b88d85b9f8?ref=runtimewire) identifies Liang as a Guangdong native who studied electrical and electronic engineering at Zhejiang University and founded High-Flyer Quantitative Investment Management in 2015. AP also reported that High-Flyer had accumulated 10,000 Nvidia A100 GPUs by 2022. [High-Flyer's account of its history](https://www.high-flyer.cn/en/fund/?ref=runtimewire) says the firm began exploring automated quantitative trading with machine learning in 2008, executed its first deep-learning-generated trade on October 21st, 2016, and later built the Fire-Flyer computing platform. High-Flyer says the platform supports AI research, while its asset-management operation works with more than 10 petabytes of data and serves over 10,000 high-net-worth individuals and institutional investors. Liang has described curiosity as his motivation for starting DeepSeek. "People may think there's some hidden business logic behind this, but it's mainly driven by curiosity," he told Chinese technology outlet Waves in a passage [retrieved and reported by AP](https://apnews.com/article/deepseek-founder-liang-wenfeng-china-ai-0673d5c39d90108189cc31b88d85b9f8?ref=runtimewire). AP also reported Liang's argument that Chinese AI teams needed to produce original research rather than continue following work from elsewhere. DeepSeek initially drew on High-Flyer's capital and infrastructure, according to the research reviewed for this article. Outside financing was reported in 2026, although DeepSeek has not published final terms. [Axios reported in June](https://www.axios.com/2026/06/03/china-deepseek-billion?ref=runtimewire) that DeepSeek was raising about $7.4 billion at an approximately $52 billion valuation from participants including Tencent, battery maker CATL, Liang and a state-backed fund. DeepSeek has not publicly confirmed a final round size, valuation, structure or complete investor list. ### DeepSeek distributes weights and sells API access DeepSeek differs from OpenAI and Anthropic in how it distributes important model releases. The [DeepSeek-R1 repository](https://github.com/deepseek-ai/DeepSeek-R1?ref=runtimewire) provides model weights and code under the MIT license, which permits commercial use and modification. DeepSeek's [V4 preview](https://deepseek.com/en/news/v4-preview/?ref=runtimewire) also introduced open-weight Pro and Flash variants with one-million-token context windows. V4-Pro has 1.6 trillion total parameters and 49 billion active parameters, while V4-Flash has 284 billion total parameters and 13 billion active parameters. DeepSeek also sells managed access through its [official API platform](https://www.deepseek.com/platform/?ref=runtimewire). That combination gives developers the choice of using DeepSeek-hosted inference or operating released models through infrastructure controlled by other parties. OpenAI and Anthropic keep their principal commercial systems closed. The distinction is relevant to the range of organizations invited to the meeting, but Reuters did not report whether the council planned to examine open-weight distribution or ask DeepSeek about it. Reuters identified the subject only as AI and international security. ### A safety assessment, with limits [Axios reported in July](https://www.axios.com/2026/07/07/report-ai-safety-pledges?ref=runtimewire) that DeepSeek received a failing overall grade in the Future of Life Institute's AI Safety Index and did not complete the institute's survey. That assessment does not establish how DeepSeek's systems perform in practice. Axios said the index was based largely on public policies, research, reporting and disclosures, supplemented by a survey sent to AI labs. Seven outside reviewers assigned grades across 37 indicators. DeepSeek's nonresponse limited the information available to those reviewers, but it does not independently establish DeepSeek's internal safety practices. Reuters reported that leaders of major Chinese model developers have rarely discussed publicly whether increasingly capable AI could deceive humans, evade control or cause catastrophic harm. DeepSeek's public materials have instead emphasized model capability, computing efficiency, open releases and a long-term pursuit of artificial general intelligence. Reuters did not report what DeepSeek intended to say about safety or governance if DeepSeek accepted the invitation. The diplomatic setting is already contentious. [Reuters reported](https://www.reuters.com/world/china/china-state-newspaper-blasts-anthropics-calls-slow-ai-cold-war-tactic-2026-09-14/?ref=runtimewire) that China's state-backed Global Times described Anthropic CEO Dario Amodei's call to slow advanced AI development as a "Cold War playbook" intended to preserve Washington's technical lead and exclude China from global AI governance. That characterization belongs to the Global Times and does not establish DeepSeek's position. Senior US and Chinese officials have also agreed to continue bilateral talks about AI dangers and communications protocols for safety incidents, [Reuters reported](https://www.reuters.com/world/asia-pacific/us-china-meet-again-ai-safety-two-months-shenzhen-bessent-says-2026-09-21/?ref=runtimewire). The Security Council meeting brings the discussion into a 15-member multilateral forum. The council [first met specifically on AI risks in July 2023](https://www.reuters.com/technology/un-security-council-meets-first-time-ai-risks-2023-07-18/?ref=runtimewire). China warned then that AI should not become a "runaway horse," while the United States focused on the risk that governments could use the technology for censorship and repression. For September 23rd, the confirmed facts remain narrower than the meeting's potential implications. Altman planned to brief the council, and Anthropic's presence was expected but unconfirmed. DeepSeek and Moonshot were invited to make statements. Whether either accepts, who speaks and what position is presented remained unresolved on September 22nd.