Bloomberg Audio Studios, podcasts, radio news.
If you were in Shenzhen in early March, you might have caught an unusual sight outside the headquarters of one of China's biggest tech.
Companies, Chanjo more say the open Claw.
Crowds of people huddled around mixshift tables in an open courtyard. Everyone had a laptop in hand. People were shoulder to shoulder typing, tinkering, troubleshooting. It looked like an outdoor hackathon had broken out in the middle of the city.
There were hundreds of people from retire's elderlies to school kids and of course Yawn office workers.
LUs Ding covers China Tech for Bloomberg out of Hong Kong.
And they were all there to get tension engineers to install open Claw on their computer.
Loose what is open claw.
It's a very capable AI agentic tool. It's much more capable than AI chatbots. It can't help you to do your work on your computer like swarming files, edit videos, making purchases.
Tools like open Claw aren't something you can just download from an app store.
They require technical setup and a bit of know how.
That's why in China people are organizing gatherings like the one at ten Cents headquarters to help potential users get open Claw up and running. Nearly a thousand people showed up to that particular meetup, one of many across the country this past month.
I think it really is a vouch for the amount of AI enthusiasm that we've seen in China that we really haven't seen in other places.
Catherine Thorbeck is a Bloomberg Opinion tech columnist based in Tokyo.
When you have local governments coming out and signaling support and sort of the big tech players and the small startups and the model builders all jumping on this bandwagon at once, I think it really created this sort of title wave of you know, it just shows how fast technology trends can really spread China.
China has the most open claw users in the world, almost double the activity in the US. That's according to cybersecurity firm Security Scorecard. But as the hype surrounding open Class surged, so has the scrutiny. The initial enthusiasm for the technology is now being met with caution, as users in China's government way whether the benefits of the tool justify the risk of such deep access.
When you give these tools access to your personal data, your personal files, your computer, and then you also give them permission to take action. The amount of things that can go wrong really just compounds. So there have been reports of these tools sort of going rogue and you know, deleting all of your emails or deleting all of sort of your personal files.
This is the Big Take Asia from Bloomberg News. I'm Wanha.
Every week we take you inside some of the world's biggest and most powerful economies and the markets, tycoons and businesses that drive this.
Ever shifting region.
Today in the show, China's rushed toward open Claw. What the technology does, why it took off so fast in the country, and what it tells us about the risks and realities of AI's next base. When open Claw launched in November of last year, it took off right away. Part of that success was the product itself. Open Clau's
founder describes it as AI that actually does things. One of the tech world's biggest names, Jensen Huang of Nvidia, called it the single most important release of software probably ever. The other part of that success the popularity of the lobster.
The logo open claw looks like a very cute cartoonish lobster. It looks like Aminion from the Menims movie, but then it's read and it has claws. Also spicy lobster. It's very favorite dish in China. Part of the fan fair is also because of that.
Spend five minutes at an open claw event and you'll see lobster hats, lobster plushies, and you'll hear a phrase that's become popular in China, yng loncha, which roughly translates to raising lobsters loose? What is raising lobsters?
Raising means that you first install it and then you spend time with it, and of course spend money on it, and then open clall learns what you like and dislike. Say, it learns different skills, and it's like raising a pat, only the pet is a virtual AI assistant that can work for you.
A virtual AI assistant also known as an AI agent. These agents are based on large language models from open ai and Anthropic, but you don't just chat with them. You can figure them to do things for you. Come through emails, book flights, and appointments. They can even trade crypto or negotiate a deal for car for you, depending on how much access you granted. And while Openclaw is hugely popular in China, it isn't a Chinese product. It
was created by Peter Steinberger, a developer from Austria. But since the AI tool is open source, it's free for anyone to use or modify. Earlier this year, Chinese developers started tinkering with it and it took on a life of its own in China.
Since after China came back from the Lunar New Year holiday, there were more discussions about it on social media and then a lot of people wanted to try it. It's a very new tool that promises a lot of things that the previous AI tools cannot do.
In China, AI is very much a part of daily life right now. The country's most popular AI app, Gobao, has roughly two hundred twenty seven million active users every month. That's almost double that of deep Sicks.
But even in AI.
Hyped China, the popularity of open Class stands out.
We live in a very competitive society in China, and starting from kids, we were told that you need to always be at the front. You need to do your best all the time, otherwise you will leg behind. And then with there's something new and then people's reactions are usually just to jump on it before everybody else and
not be late. For example, I talked to this computer engineer in Shinjin who after starting using AI agents, he's been using it all the time, and he was telling me that he couldn't even go to sleep without giving some task for his AI to do. And then one of our colleagues in China, she talked to this office worker whose boss told her to use AI otherwise you
get fired. Really, the bosses are anxious as well, because they also want to have their employees to be very productive with AI and then to make more money.
Part of that anxiety comes from pressure at the top. In China, the push for AI is coming from the highest levels of the government, and AI even made its way into President Siegenping's New Year address this year.
Curzy s many large AI models have been competing in a race to the top, and breakthroughs have been achieved in the research and development of our own chips. All this has turned China into one of the economies with the fastest growing innovation capability.
Bloomberg Opinions Katherine Thorbeck says the government is aggressively driving this AI push.
The government really wants to be a world leader when it comes to AI technology, and when you look at the AI plus plan that they rolled out last year, they really are trying to use AI to solve a lot of sort of societal problems, whether that's boosting consumption, which it's not totally clear how AI will do that, but they definitely are trying to integrate AI into as many facets of society as possible. And then another reason I think that it's moving so fast in China is
because of the open source embrace. All of the big Chinese AI model builders are really going full steam ahead on open source, which means they're putting out their new models, they're putting out the weights for them, they're free for people to use and download. And I think as a result of that, this whole ecosystem of developers and tinkers can all sort of look at the code and move things along a lot faster.
In other words, when AI becomes a political priority, tech trends like open claw can move quickly. Several local governments have already backed that momentum by offering more than a million dollars in subsidies to support open claw and similar open source projects, and the country's tech giants have followed suit. Ten Cent and buy Do launch their own versions of open claw in March, using them as a way to pull people onto their page platforms. Loose tried some of them out for herself.
Now there are a lot of cloud based offers that you can raise the observer on the cloud so it doesn't access your local desktop and local files. I chose two platforms. One is from ali Baba and it was a free service. Another two that's under by Dance. It's TikTok's parent company. I paid ninety nine per month and that's about fourteen to fifteen dollars. Once you paid for a month, you have a certain amount of tokens or
credits that you can use. If you, for example, wanted to start your email, you need to granted the access to your email and then it will do it.
And what did you ask it to do for you?
I asked it to send news briefs about the companies that I cover as a journalist. It did not do a very good job because the news resources that it can access is very limited. And now it just asked it to send me a cat photo every hour.
A cat photo every hour.
Yeah, that's what I need in my life.
But beneath the novelty, there's a serious difference between open claw and traditional AI chat bots like deep seek and chat GBT, and that's the amount of access these tools require.
For open Claw, you need to give it a lot of access. For example, it needs to read your files to swart your file. It need to access your editing software to edit for you. And if you wanted to buy things for you, you need to give it to your credits information, which is kind of a redfleg.
Beijing may be all in on the idea of spurring innovation with AI, but open claus powerful capabilities are giving the government pause after the break.
Why the sweeping access.
Granted to AI agents makes Chinese regulators uneasy, and why some Chinese users are shying away from it too. While open Claw has been embraced by users, tech giants, and even some local governments in China, the response from Beijing has been more restrained.
Open Clay open there.
The Chinese government has issued a security alert on the risks of using open Claw, emphasizing that users might be susceptible to cyber attacks and data leaks. Bloomberg's loose Ding and Bloomberg opinions. Katherine Thorbeck say regulators have allowed the tool to spread, but with clear limits.
The military that's overseeing technology industry has issued warnings about the security risk of this tool. State media also posted articles about how to use open claw safely. Then they were also authority spanning the use of open claw in sensitive industries like government agencies, banks.
So I think the government is very aware of the risks, and I think they're just trying to strike that balance between embracing innovation and seeing where this takes them. And also they're obviously very concerned about sort of data leaks and data privacy.
That caution centers on one big question, what happens when these tools are given broad access to people's data and devices. You mentioned the security risks associated with open claw, What are those and what can go wrong?
Oh gosh, So when it comes to the cybersecurity risks for open claw, I'm not even sure where to start. I think when you give these tools access to your personal data, your personal files, your computer, and then you also give them permission to take action, the amount of things that can go wrong really just compounds. So there have been reports of these tools sort of going rogue and you know, deleting all of your emails or deleting
all of sort of your personal files. So that's one side of it, and then on the other side, I think it also sort of invites all sorts of new sort of cybersecurity attackers and cyber attackers. And so the way I sort of think about it is, if you remember, like the two thousands Internet, when we were all just kind of downloading various softwares onto our PC, and none of us really knew what we were downloading, but there was definitely a curiosity, but we might not have understood
that we're also downloading all these viruses. So I think we're seeing some of that as well with people, because, as I mentioned, you need some little bit of a tech background to really use open claw and plug ins and sort of the things that you can do on top of it. So I think people are downloading all kinds of things and not totally aware of what they're installing and what sort of permissions they're granting.
By now people in China have had some time to play around with open claw, I want to talk about how Chinese users are feeling about the tool. Now you know, after the initial excitement and frenzy, there have you heard about this one lobster freeway saying that's trending on Chinese social media?
Yes, this sort of hashtag that you're referring to that has taken off now. One lobster three ways refers to paying somebody to install open claw for them, and then paying somebody else or a lot of times that same person to uninstall open claw for them, and then also paying them to clean up the mess that open Claw might have left behind on their computer, whether that's deleting files or swing all kinds of chaos. So I definitely think that the initial excitement has moderated a little bit.
But even as the hype settles, Catherine says, it's clear that China's big tech companies are betting on agenic Ai to be the next big driver of growth. The country is a testing ground for it, and what happens next is still worth watching.
I really look at it as this real world experiment where regular people are playing around with agentic tools and seeing what it can do for their lives, and it can actually boost productivity and sort of what the risks are, and that's happening at such a massive scale in China, and it's really a double edged sword because the more permissions and data and access that you give to AI agents, the more they can do, and that's really exciting. But then at the same time, the flip side of that
is the more damage they can also cause. So I think making AI agents secure isn't something that either side has really figured out yet, and I think it's going to take a lot of time. I will say that one benefit to open claw is that it is open source, so that means that the more people at this scale that are all using it at the same time, the more it sort of hardens it because people are, you know,
figuring out bugs, then people are sending patches. I think if you look at it at the enterprise level, this has the potential to be absolutely enormous, and I think in ways that we don't even fully understand, you know. I think if you can fully automate the work that say an intern does or an entry level worker does, I think that can really have profound effects on the labor market.
And in terms of the flip side, where could this experiment fail.
I think in the longer term, you know, as much as the tech industry likes to say that AI agents are there to empower workers. I don't necessarily buy that.
I think the goal is really to replace workers, and so I think in China that's actually going to be a really big thing to watch because youth unemployment has remained persistently high since the pandemic, and I think if you have large swaths of educated and unemployed young people that are looking for jobs and that are being put out of work by these AI tools, I think that doesn't bode well for social stability, and I think that will be sort of a big test for the Chinese government.
I think in the near term we'll have to see how much this actually sticks and if people are actually able to build trust with AI agents.
And ultimately, Catherine says, how China handles tools like Openclaw could shape the next phase of the global AI race.
Building the top of the line frontier models is side of it, and then the other side of the sort of AI race is diffusion and how do you integrate AI into as many facets of the economy and as many facets of society as possible. And I think we're
seeing China really pick up on the diffusion race. We're really seeing that they're willing to try out this new technology, willing to see how it can increase productivity in their personal lives and their work lives, which I think actually is the race that we should be watching, because I think that's where you really see the biggest benefits from major tech evolutions in history. I think from that perspective, it definitely shows that China is sort of leaping ahead
when it comes to AI. They're embracing it, they're using it, they're experimenting, and I think in doing so, they're also sort of making their own models better. You know, they're giving real world feedback to their own models, and I think that just only improves them. So I think that could really tip the scales in the long run when it comes to this broader sort of AI race.
This is The Big Take Asia from Bloomberg News. I'm wanha.
To get more from The Big Take and unlimited access to all of Bloomberg dot Com, subscribe today at Bloomberg dot com slash podcast Offer. If you like the episode, make sure to subscribe and review The Big Take Asia wherever you listen to podcasts.
It really helps people find the show. Thanks for listening, See you next time.