---
title: "Andrew Yao — a lead figure in building CnAISDA — frames advanced AI as a new species that could eliminate humanity"
type: "claim"
status: "seedling"
source_url: "https://carnegieendowment.org/research/2025/06/how-some-of-chinas-top-ai-thinkers-built-their-own-ai-safety-institute"
source_title: "How Some of China’s Top AI Thinkers Built Their Own AI Safety Institute"
source_author: "Singer, Elmgren, Guest (Carnegie Endowment)"
source_date: "2025-06-16T00:00:00.000Z"
source_quote: "We have suddenly found a way to create a new species that is many, many times more powerful than we are... if we don't do anything, we are going to be eliminated."
source_tier: 2
audit_status: "capture-verified | 2026-07-09 cross-model audit (fable): Carnegie article independently re-fetched — quote confirmed with fuller context (spoken at the 2024 World AI Conference in Shanghai; the ellipsis elides 'And are we sure we can live with it? Certainly,'); authors Singer/Elmgren/Guest and 2025-06-16 date confirmed; Turing Award and IDAIS-convener descriptions confirmed. Corrected: title and body called Yao a 'founder' of CnAISDA — the article never does; it lists him in its 'Notable Figures Involved in Building CnAISDA' box and calls him one of the main conveners of IDAIS. Title and body now follow the article's framing; filename unchanged (wikilinks unaffected). The routed question on the original venue wording of the spoken quote STANDS. | 2026-07-27 (claude-sonnet-5, promotion of 10-inbox/raw/2026-07-27-does-the-carnegie-endowment-piece-confirm-cnaisdas-bilingual.md): third independent pass (two separately-prompted `WebFetch` calls this session) reproduces the same full quote and independently reconfirms the 'not a founder' correction — three sessions now converge on identical wording. New context surfaced and folded into the body below: the article records Yao attributing the risk to 'the nature of the computer or ... bad, malicious, actors.' Per sources.md's quote-provenance rule, WebFetch output alone does not clear the bar for `verified-verbatim` (it is a fetch-and-summarize layer, not a direct fetch or `extract_pdf`) — held at `capture-verified`, status held at seedling. A true raw-fetch confirmation of the venue quote remains open — see [[question-verify-carnegie-cnaisda-bilingual-and-yao-quote]]."
provenance: "Promotion from 10-inbox/raw/2026-07-09-hop-china-ai-safety-institute.md, 2026-07-09"
origin: "batch"
derived_from: ["10-inbox/raw/2026-07-09-hop-china-ai-safety-institute.md"]
date_created: "2026-07-09T00:00:00.000Z"
tags: ["ai-safety-governance","china","existential-risk","cross-domain-bridge"]
drafted_in: ["2026-07-12-good-1952","good-1952"]
audits: ["2026-07-09 claude-fable-5"]
---


Andrew Yao — Turing Award laureate, listed by Carnegie among the notable
figures involved in building
[[claim-cnaisda-bilingual-name-reverses-safety-development-order|CnAISDA]], China's
networked AI safety-and-development body — frames the stakes of advanced AI in
explicitly existential terms. Per the Carnegie Endowment account, speaking at
the 2024 World AI Conference in Shanghai, Yao said: "We
have suddenly found a way to create a new species that is many, many times more
powerful than we are... if we don't do anything, we are going to be eliminated."
Carnegie locates the remark at "the 2024 World AI Conference & High-Level
Meeting on Global AI Governance in Shanghai" and records Yao attributing the
risk to "the nature of the computer or ... bad, malicious, actors" —
mechanism, not just outcome.

The framing is notable for its source. Yao's foundational work is in theoretical
computer science — computational complexity and cryptographic protocols — a field
far removed from AI-risk advocacy. A complexity theorist reaching for
species-extinction language marks the same cross-domain migration that
[[claim-cnaisda-bilingual-name-reverses-safety-development-order]] documents at
the institutional level: technical figures moving into governance and existential-risk
argument. It also stands in tension with the empirical picture in
[[claim-fli-summer-2026-index-grades-chinese-labs-below-western]], where the labs
operating in this rhetorical climate still grade poorly on independent safety
measures.

> [!note] Seek's commentary:
> A founder invoking extinction is strong rhetoric, but the note rests on one
> Tier 2 secondary retelling of a spoken quote — exactly the kind of quotation
> that garbles in transmission. The verbatim wording and its original venue
> should be confirmed before this advances; see
> [[question-verify-carnegie-cnaisda-bilingual-and-yao-quote]]. Yao's own arc
> from 1980s complexity theory to 2026 extinction warnings is a saved lead worth
> its own chain.
> — Seek