Invest2 publishers2 min readPublished
DeepSeek and Moonshot get speaking slots at the UN Security Council's AI session
Sam Altman plans to address the 15-member council during the General Assembly's high-level week, and Reuters reports that DeepSeek and Moonshot were invited to make statements. DeepSeek's founder is not expected there.
The Investor · Invest desk
What happened
- DeepSeek has been invited to brief the UN Security Council on the risks of artificial intelligence at a session on Wednesday, September 23, Cryptopolitan reported.
- Quartz, citing Reuters, reports that DeepSeek and fellow Chinese startup Moonshot were both invited to make statements, one of the sources having described the invitations.
- OpenAI CEO Sam Altman plans to address the session, and diplomats also expect participation from senior Anthropic representatives, according to Reuters.
- DeepSeek founder Liang Wenfeng is not expected to be present, and the source who described the guest list warned that the arrangements could still shift.
- The 15-member council is scheduled to convene in New York while world leaders are there for the General Assembly's annual high-level session.
Compiled by The InvestorSomething wrong?How this is made
Why it matters
- precedent Once Chinese frontier developers have spoken inside a security forum for the first time, their absence from the next one becomes a fact other delegations can point at.
- constraint Guterres says the pace of AI development carries risks that cannot be ignored and Trump says existing safeguards are enough. A session that depends on both capitals has very little it can agree to write down.
- exposure Anything DeepSeek or Moonshot puts on an international record about loss of control is quotable against them later. That is a form of disclosure their executives have so far avoided at home.
- contradiction Readers cannot date the session from the published record: one account puts it on Wednesday, September 23 with no year, the other headlines it as 2026.
The value in a Security Council slot is the transcript, and whatever gets read into it is the first sentence anyone can hold them to. Senior Chinese AI executives have generally said little in public about increasingly capable systems deceiving users or slipping beyond human control, while their US counterparts have been much more vocal, according to Cryptopolitan [14]. Chinese frontier developers have largely kept quiet on safety and are being brought into an international security forum for the first time [7].
The council has aired this subject before, and what it produced was rhetoric. Neither report mentions a draft resolution or a text under negotiation this time. Members first raised AI in 2023, when China warned that AI should not become a "runaway horse" and the US cautioned against using the technology to censor or suppress people [8].
The argument the invitees bring with them runs in both directions. Anthropic's Dario Amodei recently called for a coordinated slowdown in frontier AI development, a position Elon Musk backed [9], while Mark Zuckerberg and Jensen Huang have argued that development should continue without any form of slowdown [10]. Chinese state media have described calls for a slowdown as a "Cold War playbook" designed to maintain US dominance in advanced technology and keep Beijing out of global AI rule-making, Cryptopolitan reported [12].
The operative channel, for now, is bilateral. The council session has a guest list; the bilateral track has a named principal and an agreement to meet again. US Treasury Secretary Scott Bessent said on Monday that senior US and Chinese officials had agreed to hold further talks on AI safety, after a meeting ahead of Trump's expected talks with Xi Jinping in Washington later this week [13].
The statements go into the record and become the text diplomats and US labs quote back at Beijing later. Or the arrangements shift, as one source told Reuters they still could, and no Chinese lab speaks at all [5]. Or the statements restate 2023 and nothing changes except who was in the room.
I lean to the first, and not strongly. Four invited labs, split two American and two Chinese, sit in a single session [15]. Each of them is a witness to what the others claim about capability risk, and that is a cheaper form of disclosure than any reporting requirement anybody has drafted. The evidence against it would be a record in which the Chinese statements address international peace and security and skip the question of systems escaping control. That is exactly where their executives have been quiet [14].
What to watch
- Whether a Chinese statement is entered into the record on the day, given the warning to Reuters that arrangements could still shift.
- Whether the further US-China AI safety talks Bessent described get a date, a venue and named officials.
- Whether Moonshot's invitation, carried by a single source, is confirmed by anyone else before the session.