Making sense of the horror of Chinese AI

The launch of the latest AI model from a Chinese company – Moonshot AI’s Kimi – reignited debates about American competition and open versus proprietary AI.

While there has been a lot of discussion on social media, it seems that the debate is happening behind the scenes in Washington, DC, where OpenAI and Anthropic have reportedly lobbied regulators with concerns about China’s open models.

In a recent episode of TechCrunch’s Equity podcast, Kirsten Korosec, Sean O’Kane, and I discussed why this seems to be such a hot button issue. Besides suggesting that some people should “touch the grass” rather than spend their weekends arguing about X, Sean noted that in many ways, this “feels like we’re seeing a repeat of the previous outrage,” with everyone in Silicon Valley “expecting something to come along and blow everything up.”

And Kirsten noted that imposing strict limits on China’s AI models would greatly benefit a few companies: “Do we speed it up and make sure that the Americans win the AI ​​race, or do we make sure that some frontier labs do better than others?”

Read on for part of our interview, edited for length and clarity.

Anthony Ha: For people who have followed the talk around Chinese AI, this will probably be very familiar since the launch of DeepSeek, which is basically where the Chinese model comes from; in some benchmarks, it does, or at least appears competitive with other borderline models; and part of the tech industry is losing their minds.

Some of these [debate] you got more scrutiny because one of the people was posting about it [an executive] at OpenAI. But usually, there [is] the question that arises is: Can Chinese companies beat US companies, at least in some aspects, and do it more cheaply and more transparently?

Sean O’Kane: Yes, there are many aspects of this that feel like we’re seeing repeats of previous freakouts. I think one of my favorites is: Everybody’s ready [for] so we expect something to come along and blow everything else away. And I think my favorite example of that this past week was people showing that “My gosh, Kimi did in 30 minutes a complete replication of macOS.” And yes, it made a pretty impressive reproduction of the macOS look, but it’s not an OS.

We keep seeing these things happen over and over again, when everyone is jumping around in the tech industry. And I think especially, with some of the Chinese models that come out, there’s this expectation, and I think this gets to the heart of why people reacted the way they did last weekend. (Also, by the way: Go outside, touch the grass, it’s the weekend. Everyone in the industry was trading barbs on Twitter all weekend.) But this jump is really interesting to me because we’re now one week in and I don’t think anyone feels like the end is as close as they were last week.

Kirsten Korosec: We have a really good story about one of our reporters, Tim Fernholz, who is trying to expose mental illness here in the United States. He gives many reasons. And I conclude – and I don’t want to conclude him, but I think that some rise more than others.

There are concerns that these open-weight models of China may have an inherent bias towards China, and there are other concerns about security risks and safeguards. But there is a great idea here, which is to protect, and who is going to quote-unquote “win the race”? Will it be the US or China? And that seems to drive a lot of that fear.

I don’t know, Anthony, if you agree with that?

Anthony: I completely agree. I think the China factor always adds to this certain level of hysteria. And that doesn’t mean people shouldn’t worry about how the US stacks up against China in all different industries. But it goes up a lot.

Another thing this reminds me of is the discussion surrounding TikTok a few years ago. And again, it’s not that I thought the anxiety around TikTok was completely manufactured, but the level of how panicked people are – it seems that as soon as you add the word China to any conversation, things escalate. And then in this case, it’s connected to this discussion about opening [weights] and this idea that AI is so powerful and dangerous that the only way we can control it is through these proprietary models from these American border companies.

Obviously, many people say this [have] the reasons why they want to say that. David Sacks, the Trump administration’s former AI chief [and] now he has a different role in the Trump administration, he was yelling at X about how, “I can’t believe people are against data centers, we’re tying them in knots, there are too many regulations.” So it’s a way to challenge the positions they already had about AI. “My Lord, if China defeats us, that is unthinkable, so you must do what I want to do anyway.”

Kirsten: Okay, and if you were to put a ban across the board on Chinese open weight models – I’m not saying there’s no real concern here, but let’s play that out. If we can do that, it will benefit the models created by OpenAI, for example, and it will force businesses to use those who are against using models like Kimi.

So you really have to ask the question: Are we accelerating and making sure that Americans win the AI ​​race, or are we making sure that some frontier labs do better than others?

Sean: For now, we have to say that a lot of this discussion was started by OpenAI’s head of future strategy, Dean Ball, who was the first to come out with this really long post addressing some of these concerns.

Part of me thinks the reaction to this was because people disagreed with what Dean wrote. Part of me also thinks that the reaction was because he was just saying this out loud. He said that basically the US should create regulatory FUD – fear, uncertainty, and doubt – and eliminate the power of these open weight models to compete with the US. [Ball later backed away from this argument.]

And for me, I think you can learn from some of the responses from people, like, “You shouldn’t say that out loud, Dean.”