The letter is a two-page communication dated September 16, 2026 from Representative Ro Khanna (D-California), writing as Ranking Member of the House Select Committee on the Chinese Communist Party, to Eddie Wu, chief executive officer, and Joe Tsai, chairman, of Alibaba Group Holding Limited in Hangzhou. It urges "China's leading artificial intelligence (AI) companies to enter into a binding, enforceable, and verifiable agreement alongside America's leading artificial intelligence (AI) labs to pace the frontier," and asks four written questions, with answers requested "by no later than September 22, 2026."
The Washington Post published the letter with an exclusive report on September 17, 2026, describing it as "the first overture of its kind" and reporting that Khanna was asking leading Chinese AI labs to cooperate on pacing frontier AI (Source: washingtonpost.com). The Post's account, which is the basis for reports that the letters went to the chief executives of Alibaba, Moonshot AI and DeepSeek and were sent on September 17, is a secondary description; the one letter held is addressed to Alibaba alone and carries the date September 16. The Moonshot AI and DeepSeek letters have not been located, and their text is not assumed to match.
The request
The letter opens with a claim about the US laboratories: "The CEOs of Anthropic, OpenAI, and xAI do not agree on much, but earlier this month, they all came together and agreed that we should slow the rate of AI advancement until society can develop the necessary guardrails. It is time for Moonshot, Alibaba, and DeepSeek to do the same." The statements it refers to followed Dario Amodei's "We Must Pace the Frontier" of September 12, 2026, which Elon Musk and Sam Altman endorsed in posts on X the same day; neither endorsement specified a pace or an oversight arrangement the endorser would accept (We Must Pace the Frontier (Amodei, September 2026)). Later in the letter Khanna describes these statements as the founders' "call to pace the frontier" and a "pledge" that the Chinese companies should "echo."
It then rejects the argument that "any strategic slowdown in AI development necessarily means ceding American AI leadership to Beijing," calling that view "distinctly misguided, in part because it presumes global pacing is not a possibility." Its premise is a shared interest: "the United States and the People's Republic of China (PRC) share a mutual interest in preventing the loss of control or misalignment of advanced AI systems."
Two tracks: governments and companies
The letter sets out a government track and a company track.
On the government side, Khanna states that he has called on the governments of the United States, its democratic allies and the PRC "to urgently engage in negotiations to establish such baseline global standards." To illustrate the shared risks the letter quotes Amodei's essay — "worry that in 6–12 months a swarm [of agents] could be capable of taking over the entire internet with a persistent botnet" — and attributes to Altman the statement that "we could lose control of the future to AI," sourced in a footnote to a post on X. It adds that neither country "would benefit from a biological or chemical attack facilitated by AI," and calls for leaders to "negotiate an enforceable U.S.-PRC AI Treaty," with discussions to "commence at the meeting between President Trump and General Secretary Xi next week." It states: "Unless the world adopts verifiable frontier pacing standards, including explicit limits on unchecked recursive self-improvement, I fear that this powerful technology will not always remain in service of humanity."
The company track is framed as a fallback: "if our political leaders do not pursue or fail to achieve such an agreement, it falls on leading AI companies themselves – wherever they may be domiciled – to ensure their decisions do not jeopardize the collective fate of humanity." The letter argues that if all leading laboratories agreed to pace the frontier, "the economic incentive to prioritize profit over safety would be rebalanced," and governments would gain time to negotiate a binding agreement.
Acknowledged differences
The letter does not present cooperation as rapprochement. It states that the PRC government "commits grave human rights violations against its people, and PRC AI models have played a significant role in facilitating those abuses," that PRC models are used "to the detriment of American and allied interests, including by helping to modernize the People's Liberation Army and supporting PRC intelligence agencies," and that "Chinese law can compel such cooperation." It limits the proposed cooperation to avoiding "practices that could result in humans losing control of AI or the development of rogue misaligned agent swarms," on the reasoning that simultaneous commitments by both countries' laboratories make follow-through more likely: "We do not need to agree on much to agree that we should narrowly work together on this."
The hearing and the questions
Khanna writes that at "a congressional hearing I am convening next Wednesday" (September 23, 2026, counting from the letter's date) he will urge President Trump and General Secretary Xi to pursue AI diplomacy and "demand that your company and the other leading Chinese AI labs adopt frontier pacing alongside your American peers." The Post described it as an emergency hearing on international mechanisms to pace AI (Source: washingtonpost.com). The hearing had been announced, not held, as of September 18, 2026.
The four questions ask:
- Whether the company will agree to "pace the frontier" and slow its rate of AI advancement so that societal safeguards can catch up.
- Whether, if the leading US laboratories offer "a binding, enforceable, and verifiable agreement" to pace the frontier, "including by adopting common sense safeguards like global limits on unchecked recursive self-improvement," the company would execute it.
- Whether, under such an agreement, the company would accept "neutral third party auditing and all-access inspections to ensure verifiable compliance."
- If it would not agree to any of these commitments, what safeguards it has "to prevent the loss of control or misalignment of your AI models and future AI models."
Relation to other pacing proposals
The letter addresses foreign developers directly, where other September 2026 proposals addressed governments or US firms. Amodei's essay places attempted coordination with authoritarian governments as its third step and treats an international "speed limit" on recursive self-improvement as difficult but possible (We Must Pace the Frontier (Amodei, September 2026)); the letter names limits on "unchecked recursive self-improvement" as a component of both the proposed treaty and the company agreement. The Center for American Progress had called for the United States and China to move toward a government-level agreement on pacing frontier releases, a position that became public on September 15 (Source: insideaipolicy.com), and Musk had proposed on September 14 that "three or four of the leading Chinese companies" join US developers in letting rivals test their models before release (Source: cnbc.com). The letter's third question specifies a verification form, third-party auditing with "all-access inspections," that goes beyond the embedded evaluators in Amodei's first step. See Coordinated slowdown proposals compared.
Context and reception
The Post's account states that Chinese officials and technology leaders had so far rejected calls to slow AI development (Source: washingtonpost.com). China's Foreign Ministry had that week pushed back on US AI companies' calls to slow down and warned against "fear-mongering" (Source: axios.com). On September 17, Huawei rotating chair Eric Xu urged Chinese AI companies to accelerate development to catch up with frontier US laboratories, saying Chinese researchers needed to "increase the speed of development so they can also see the dangers of AI development" (Source: ft.com). No response from Alibaba, Moonshot AI or DeepSeek had been reported as of September 18, 2026.
Provenance
The text is taken from a PDF in the Washington Post's document repository, linked from its September 17 report. No copy was located on Khanna's House site or the select committee's Democratic site, so the capture rests on a secondary host; the Post's report quotes two passages that appear verbatim in the document. Footnote 1 cites Amodei's essay at darioamodei.com; the letter's quotation replaces the essay's "such a swarm," which refers to a more capable swarm with a similar level of misalignment to the one in the OpenAI–Hugging Face incident, with "a swarm [of agents]." Footnote 2 cites an X post (status 2099352016988614852) for the Altman quotation, which has not been independently captured.
Relationships
- related: Coordinated slowdown proposals compared — the letter as a congressional request to foreign developers to join a pacing agreement
- depends-on: We Must Pace the Frontier (Amodei, September 2026) — quoted in footnote 1; the US statements the letter asks Chinese laboratories to echo
- related: US-China AI Competition: Different Races, Different Metrics — rejects the argument that pacing cedes leadership to Beijing; proposes a U.S.-PRC AI treaty
- related: Recursive Self-Improvement (RSI) — asks for "global limits on unchecked recursive self-improvement"
- related: Ro Khanna, House Select Committee on the Strategic Competition between the United States and the Chinese Communist Party — sender and the committee in whose capacity he writes
- related: Alibaba / Qwen Team, Moonshot AI, DeepSeek — addressee and the two other reported recipients
- related: Xi Jinping, Donald Trump — leaders the letter asks to begin treaty discussions at their September 2026 meeting
- related: AI Race Dynamics — the collective-action premise the letter disputes for the US–China case