Exclusive | On the Kimi Case, Allen Zhu Spoke With Us for 20,000 Words Straight
On gain and loss.

By Lili Yu
Edited by Jing Liu

Allen Zhu has finally broken his silence — and he's pointing his finger directly at Rita Zhang.
This past Thursday, Zhu posted twice in quick succession on WeChat Moments. The first was a response to the recent arbitration case involving Yang Zhilin. The second drew a sharp line through the cast of characters: he was willing to exempt Zhang Yutao and Moonshot AI, supporting young people's pursuit of AGI dreams; but he separated Rita Zhang out, accusing her of violating fiduciary duty. Finally, he expressed his bewilderment at Kimi's insistence on keeping Zhang tied to the company.
These signals have pushed the Moonshot AI financing case, which has dragged on for more than half a year, into a new phase: it seems most of the conflicts are resolvable, but the standoff between GSR Ventures and Rita Zhang remains a "dead end."
Yesterday evening, Anyong Waves conducted a lengthy interview with Allen Zhu. He gave clear, pointed answers on all the key questions — GSR's stance toward Moonshot AI, the delicate positions of the various parties in this博弈, and the inside story of his interactions with Rita Zhang, one of the two central figures in this fight. The transcript ran over 20,000 Chinese characters; we've extracted the most critical portions below.
Over the past six months, Anyong Waves has tracked the Kimi affair from start to finish:
"Allen Zhu Turns Left, Rita Zhang Turns Right"
"Exclusive | Kimi Founder Yang Zhilin Sued for Arbitration by Former Investors"
"Exclusive | The Full Underwater Story of the 'Kimi Yang Zhilin Arbitration Case'"
We have no intention of blaming or judging any of the individuals, institutions, or companies involved. But the Kimi case has exposed a problem rarely discussed in the venture capital industry: in China's business context, how does one leave a company gracefully, and how should the stakeholders conduct themselves?
Years from now, the Kimi affair may well become a classic case study in Chinese venture capital history.

On "Decoupling": "Why tie a simple problem together with a complicated one?"
Anyong: In your two posts yesterday, you mentioned being willing to exempt Zhang Yutao and Moonshot AI. Does this mean the arbitration cases previously filed against them will be settled?
Allen Zhu: If they're willing to decouple, the previous arbitration cases can be settled immediately. Both sides have just filed — we haven't even paid the fees yet, so it hasn't officially started.
Anyong: What exactly does decoupling mean?
Allen Zhu: It means Rita Zhang's situation is separate from Yang Zhilin and Zhang Yutao's situation with Moonshot AI. We discuss them separately. The latter is very easy to settle now. The main conflict was that during the spin-off, it was said Zhang Yutao was just going over to help part-time, but later he became full-time CTO.
Yang Zhilin had said at the time that if Zhang Yutao went full-time, he would compensate us with a certain amount of shares. He still stands by that. So if he just gives us what he promised, that's that.
But Rita Zhang's problem is completely different: she fundamentally violated fiduciary duty. Those 9 million initial free shares are entirely illegal gains.
Anyong: You seem to be directing all your fire at Rita Zhang — that was the focus of your second post. What happened in the hours between the two posts? Did negotiations break down?
Allen Zhu: We weren't negotiating today. The first post was just brief, so I wrote a more detailed follow-up. The settlement was basically almost done. The only sticking point was their refusal to separate from Rita Zhang — they insisted on bundling her in.
Anyong: From the outside, it looks like Moonshot AI is tied to Rita Zhang because fundraising is so critical for large model companies.
Allen Zhu: But honestly, fundraising isn't as important as it used to be. Besides, Yang Zhilin and Zhang Yutao are a very simple problem. Moonshot AI's shareholders require them to obtain waivers from Circular Intelligence's former shareholders by December 15th of this year. If they can't get them, or if we file suit, they have to reduce their valuation by 10%. That means returning 10% of investment amounts to investors or issuing additional shares. That's a huge penalty, so they're willing to settle with us.
But Rita Zhang's problem is a very serious matter. Why are they tying a very difficult problem together with a very simple one?
Anyong: What are Circular Intelligence's former shareholders saying? Are you all unified in your position?
Allen Zhu: Since we've filed suit, we're unified in believing this sets a very bad precedent. We don't understand why the company is protecting Rita Zhang so fiercely — she's neither a director nor an executive at Moonshot AI. We also don't understand why they can't set things right and move forward unburdened.
Anyong: What about Moonshot AI's investors? What's their attitude?
Allen Zhu: Many of Moonshot AI's investors are our friends. We've talked a lot, and they've all said to let it go. Although my understanding of large models differs from theirs, that's fine — let them go, let them develop, no problem. But with Rita Zhang, we have a responsibility to our LPs. She violated fiduciary duty and misappropriated fund assets. We can't let that go.
Anyong: If things don't develop as you expect — say, Rita Zhang and Moonshot AI maintain their current relationship — what would you consider doing?
Allen Zhu: If we can't settle, we'll definitely go to court. We have to demonstrate to our LPs that we're doing everything possible to protect fund assets.
Anyong: Some people say you intentionally misspelled Rita Zhang's name in your post (note: writing 桐 instead of 彤).
Allen Zhu: It's probably because we usually use English names internally and rarely use Chinese. Her internal WeChat ID is the pinyin initial of her middle name, y. So we generally call her y — the full pinyin is too long.

On "Concealment": "She admitted she was the anonymous founder"
Anyong: Rita Zhang previously told us she would have a transition period and would continue as an advisor and provide post-investment support at GSR. But your version is that she was fired. Which is more accurate?
Allen Zhu: The advisor arrangement was intended as an amicable separation. There were already many rumors in the market, and she denied them all, so we chose to believe her. Later she said she wanted to explore some AI opportunities, and we were fine with that. Whatever she wanted to explore, we would preserve all her interests at GSR. We were being extremely generous.
We gave her an amicable separation agreement where she could serve as an advisor. It required her to sign a commitment that she had no direct or indirect interest in Moonshot AI. But she never dared to sign it. Later we discovered she held massive shares in the company. Once we had solid proof, she had to be fired, right? The so-called advisor rehire never took effect because she never signed.
Anyong: How did the process of getting solid proof unfold? When did you get involved?
Allen Zhu: Around Qingming Festival this year. At first we also heard market rumors that Rita Zhang had interests in Moonshot AI. Later we asked in our internal company WeChat group, and she kept denying it. We asked for the cap table, and she wouldn't provide it. So we had no evidence.
But around Qingming Festival, after the previous round's cash-out controversy, some people who hadn't gotten their share of the money came to me and very kindly sent us the cap table. That's when we discovered she had 9 million free shares in the company. Later she also took over 3 million shares from Yang, totaling 12 million. She was the second-largest individual shareholder at Moonshot AI's founding, second only to Yang Zhilin.
Anyong: Whose name were these shares under?
Allen Zhu: Under her husband's BVI (British Virgin Islands) company. Regarding whether these shares belonged to her husband or to her, when we later sent a lawyer's letter to inquire, she admitted these shares were hers from day one, as a founder. Later, during Alibaba's financing round, there was also a clause requiring everyone to agree to transfer her husband's nominee-held shares to a certain anonymous founder. We asked who this anonymous person was, and her lawyer's letter admitted this BVI was backed by herself.
Anyong: What happened between the denials and the admission?
Allen Zhu: Within GSR, she didn't admit it. Later, two directors appointed by Circular Intelligence's shareholders met with her, and she also denied it, answering "go ask my husband." Later we hired lawyers — lawyers are expensive and annoying, you know? We sent three lawyer's letters, and after three rounds of inquiry, she said she had her justifications, and that she was the anonymous founder. To this day I don't quite understand why she admitted it.
Actually, even if she hadn't answered, we would have found out. Later we heard the lawyer at the time was fired. But this will all be testified to in court — other investors, directors, it's hard for them to lie for you.
Anyong: Were Moonshot AI's investors aware of this?
Allen Zhu: Because it was under a BVI company, many investors didn't know who it was. Most investors didn't put in much, their stakes were small, and they didn't know who was who. After this was exposed, some of their investors were also furious.
Anyong: So this structure was quite concealed.
Allen Zhu: Perhaps she had many misgivings about what she was doing, so she thought of many ways to cover it up. For example, when the company was valued at $200 million, she had her husband invest $150,000 for a very small percentage, so her husband could legitimately appear on the company's shareholder list. And when people questioned why this person was on the shareholder list, she could say he invested real money. But ordinary people wouldn't ask about more details, thus covering up the fact that she herself had 9 million initial shares.
But $150,000 was a drop in the bucket for Moonshot AI. Of course Yang Zhilin didn't care, because before the company was even established, financing funds had already arrived — this $150,000 was nothing. But we were shocked when we saw the cap table. Many early investors had put in tens of millions of dollars and didn't have as many shares as 9 million. Nine million shares, at today's price of 12 RMB per share, is worth $100 million.
Anyong: Here, does fiduciary duty refer to the fund where Rita Zhang was a managing partner, where related interests must be disclosed promptly?
Allen Zhu: All funds and all employees must disclose. For any project we invest in, any gains from that project belong to the fund and must count as returns for LPs.

On Fiduciary Duty: "To make it this big, this obvious, this blatant — that's really rare"
Anyong: Regarding Circular Intelligence's spin-off, you mentioned the Moonshot AI project existed as early as 2021 and was core IP developed over two years. Can you elaborate?
Allen Zhu: The Moonshot AI project team existed in 2021. Huawei's Pangu large model, for example, was done by this project team. Circular provided the people and the know-how. From 2021 reports you can see Circular Intelligence started large model training in 2021, and it was named the Moonshot Project.
Anyong: But the large models of that time were quite different from later GPT models.
Allen Zhu: They were all LLMs — perhaps just the difference between GPT-2 and GPT-3.5.
Anyong: Was the departure of "over 20 employees" agreed upon in advance, or was it later voluntary?
Allen Zhu: Agreed upon in advance. The only point of contention was Zhang Yutao. It was originally said he was going over part-time, but later became full-time. But Yang Zhilin said he would compensate according to the share ratio promised at the time — we have no disagreement on that.
Anyong: Circular Intelligence CEO Chen Qicong's role was also critical. Did he agree to the spin-off at the time?
Allen Zhu: Whether the Circular Intelligence CEO agreed or not doesn't solve the problem, because splitting off this much core assets requires all shareholders, and according to the company charter, requires majority approval from shareholders of each funding round, plus board approval.
Chen Qicong asked Rita Zhang in March and October last year whether she had joined Moonshot AI, and she denied it both times. And even if the CEO agreed, it wouldn't be enough — she still had to disclose to all shareholders and directors. Rita Zhang holding 9 million shares should also have required her to recuse herself from voting, right? That would have been a proper, legal, compliant spin-off. According to the company charter, each round of investors needed two-thirds approval to be legal and compliant.
Anyong: Some people argue that "director" isn't a full-time job, so the duties required may not be that strict.
Allen Zhu: Board members aren't paid, but that doesn't mean your obligations and responsibilities are tied to how much money you receive. The company charter is very clear about directors' obligations and responsibilities. You can't violate rules — that's the bottom line.
Anyong: Regarding fiduciary duty, this hasn't been much discussed in China's investment industry before.
Allen Zhu: In Hong Kong or Cayman, suing for fiduciary duty basically wins every time. It's the same in mainland China.
Anyong: What about USD funds?
Allen Zhu: Because previously USD structures all went through offshore entities, honestly it was harder to gather evidence. I saw many people commenting below my posts. Many said it's just rare to see something made this big, this obvious, this blatant.

On "Sour Grapes": "Talk of generational conflict, GSR not believing in large models — it's all covering up the facts"
Anyong: One narrative is that Rita Zhang proactively asked GSR whether to invest in Moonshot AI, but you declined.
Allen Zhu: We indeed never believed in it, so when they came out and asked us in March or April last year, we said no. This involves understanding of the technical model and business model. Have all my previous predictions been confirmed? To date, haven't large models failed to differentiate? So we never wanted to invest, and still don't.
But the core issue here is that she didn't disclose she had 9 million shares.
Anyong: Some people think you old shareholders, represented by yourself, have a "sour grapes" mentality, and even think your differences stem from a generational conflict.
Allen Zhu: What generational conflict? We've also invested in many young people's companies, AI companies. Generational conflict is just putting gold on your own face. I posted on Moments this morning because I was fed up with the paid commenters, really couldn't take it anymore.
And what generational conflict — this violates not only fiduciary duty, not only business ethics, but also the law.
Anyong: What do you think was the original source of all these problems?
Allen Zhu: Honestly, if Rita Zhang had been upfront from the beginning, it wouldn't have been a big problem. But she didn't disclose from the start. The longer it dragged on, the bigger the problem became.
Anyong: You've been communicating about this for at least half a year, and a settlement has remained elusive. What's the fundamental reason?
Allen Zhu: Regarding the spin-off, there's always a number everyone can agree on. The problem is the illegal gains — you have to give everyone an explanation, you have to apologize, and you have to return some of these shares. But you just won't admit it, you just hide, and paid commenters talk about generational conflict, differences between two generations, and say GSR doesn't believe in large models — it's all covering up the real facts.
Anyong: What do you think would have been the proper way to handle this from the earliest stage?
Allen Zhu: From the day you held shares, or from when you knew you held shares, you should have disclosed to all Circular shareholders. Because this is a conflict of interest. At minimum, when signing the board resolution in January this year, you should have stated there was a conflict of interest and applied for a waiver or recusal from voting, right? You can't hide, you can't conceal.
Anyong: Right now, what do you think would be an appropriate posture for Rita Zhang?
Allen Zhu: She still needs to give everyone an explanation, needs to apologize, either return some of the shares or return most of them — in any case, you have to resolve this. You can't not admit it and just hide.
Anyong: Another key question: As you said, Rita Zhang was fired. Does this mean she won't receive her Xiaohongshu carry?
Allen Zhu: We've actually been negotiating with her all along, and were very friendly to her initially. Because she kept denying she had shares in Moonshot AI, we gave her a "friendly separation package" — as long as she signed acknowledging she had no indirect or direct interest in Moonshot AI, she could preserve all her interests at GSR.
But she never signed. By the time we got solid evidence, it was the end of April and there was really no choice — we had to fire her. After being fired, all her interests are gone. When she worked on the Xiaohongshu project, she wasn't yet a partner, so it was entirely a performance bonus. The prerequisite for distributing performance bonuses is that you must be employed by the firm. If you're not a GSR employee, you can't share in the performance.
Anyong: Through this affair, do you have any deeper understanding of the LP-GP-founder chain?
Allen Zhu: Investors are the weakest link once their money is deployed. Now, the most valuable assets have been taken away, and how much to distribute is entirely up to the founders.
Anyong: Why didn't stories like this happen during the golden age of the internet? Is it because the entire VC-founder relationship has entered a frozen moment?
Allen Zhu: Back then, investors suffered and founders suffered too — which project raised hundreds of millions of USD in Series A and grew to $3 billion in a year? Which round did Wang Xing finally raise over $100 million? Many people think this is because VC relationships have hit a low point, and some even mention buybacks. But Circular Intelligence was a USD investment — it has nothing to do with buybacks, and there was no personal buyback. Human greed has always existed. In the internet golden age, there weren't such big spin-offs, so people may not have had such guts — they were more honest, more simple.
Image source: IC Photo









