Author: Shirin Ghaffary, Bloomberg

Translator: Peggy, BlockBeats

Editor's Note: This article is translated from Bloomberg's interview with OpenClaw founder Peter Steinberger. After joining OpenAI, he is involved in promoting the development of next-generation AI agent technology. The direction of enabling AI not just to answer questions but to utilize tools, collaborate across systems, and operate continuously in environments is becoming the new core of competition in the industry.

In this interview, he discusses several key questions: What do the different adoption paths of OpenClaw in China and the U.S. mean? How can AI agents be improved? How can safe collaboration be achieved between personal and work agents? How will OpenAI advance this technological direction?

The following is the original text:

The original intention of OpenClaw is to automate tasks such as flight check-ins and scheduling.

The creator of OpenClaw (who recently joined OpenAI) believes that more people should personally try using artificial intelligence to learn from it, so society can better prepare for this technology. However, before that...

You need to understand three things:

• OpenAI has stopped supporting Sora and is gradually ending its collaboration with Disney.

• Apple plans to rework Siri with AI and introduce a new interface and "Ask Siri" button in iOS 27.

• Amazon acquired Fauna Robotics, entering the consumer humanoid robot market.

Embrace AI agents

After OpenClaw gained popularity for several months, the paths taken by the U.S. and China in embracing cutting-edge artificial intelligence products have clearly diverged, and this difference may have far-reaching implications for the technological competition between the two countries.

In China, a growing number of people, from students and professionals to the elderly, are starting to try OpenClaw, with some companies even mandating that employees must use the product. Although regulators have begun to limit its application in state-owned enterprises and government institutions, China is generally becoming a large experimental field—allowing AI systems to gradually take over people's digital lives.

In contrast, in the U.S., OpenClaw (previously known as Moltbot and Clawdbot), although it has generated widespread attention among developers and early users, has not yet sparked a similar craze among the general public. Some American companies have even begun to restrict employees' use of such AI agent tools due to safety concerns.

This starkly different market response has also drawn the attention of OpenClaw's founder.

"In the U.S., I feel like in some companies, if you use OpenClaw, you might get fired," said Peter Steinberger, the tool's developer and an Austrian software engineer. He has since joined OpenAI to work on AI agent-related technology. "In China, many companies are quite the opposite—if you don’t use OpenClaw, you might get fired."

This month, at the Baidu OpenClaw "Lobster Market" event held in Beijing, there were lobster-themed peripheral products on display.

Steinberger's product was once referred to by Jensen Huang (CEO of NVIDIA) as "perhaps the most important software release ever." However, he also admitted that neither the U.S. nor China's approach is perfect. Despite OpenClaw's original intention to automate tasks like flight check-ins and scheduling, he pointed out that there are still potential security risks involved.

Note: Peter Steinberger is an Austrian software engineer and developer who gained attention for creating the open-source AI agent tool OpenClaw.

"But there's no doubt that we can learn from the faster adoption of new technologies or the acceptance of different risk preferences," Steinberger told me during an interview at OpenAI's San Francisco headquarters this week. "Ultimately, this technology is still too new; the only way we can learn is by trying it ourselves."

In his new role at OpenAI, Steinberger will be involved in developing Codex, a tool aimed at programming, which currently has over 2 million users each week. On such a highly influential platform, he is also aware that the market's demand for product safety and stability will be higher and that errors must be minimized as much as possible.

In our conversation, Steinberger discussed how to make AI agents better, OpenAI's future plans for this technology, and why he continues to keep OpenClaw as an open-source project with plans to hand it over to a forthcoming foundation under the support of his new employer. The following interview content has been moderately condensed and organized without altering the original meaning.

Original Interview

Bloomberg: Sam Altman once called you a "genius" and said you would promote the development of the next generation of personal AI agents. What will this look like at OpenAI?

Steinberger: We are rapidly moving towards a future—where everyone will have a personal agent for their private life and a work agent for their job. Through OpenClaw, I am actually building a "window to the future," showcasing the world I envision. Of course, I'm also aware that no company can truly bring it to the masses yet because some key issues still need to be resolved beforehand.

Bloomberg: What are those specific issues?

Steinberger: In that future, my agent needs to be able to communicate with your agent. For example, I work at OpenAI using Codex for knowledge work, but sometimes I need to access data in my personal "claw." So there needs to be a mechanism for my work agent to invoke my personal agent. At the same time, I also need to ensure that my personal agent does not leak any information that I consider too private, and OpenAI must ensure that internal company data is not brought back to my personal device.

Bloomberg: You have probably noticed as well that, for instance, at Meta Platforms, employees' excessive use of agent tools has caused issues, and now some companies are beginning to strengthen restrictions.

Steinberger: In the U.S., I feel that in some companies, if you use OpenClaw, you might get fired; while in China, many companies are quite the opposite—if you don’t use OpenClaw, you might get fired. They even showed me a form listing every employee's name and a column saying "What was automated today." Companies are proactively pushing employees to think about how to enhance efficiency by tenfold.

Neither approach is perfect, but we can indeed learn something from the faster adoption of new technologies and attempts with different risk preferences. Because this technology is still very new, we can only understand it through constant trials and errors.

Even at Meta, a security researcher was mocked on Twitter for publicly discussing relevant issues. I actually think that was brave. If everyone mocks these attempts, it will only discourage more people from speaking up.

Bloomberg: What do you think of the frenzy around OpenClaw in China? Many people are even lining up to experience it. Are you collaborating with Chinese companies?

Steinberger: At GTC, I've spoken with many companies like MiniMax, Kimi, and Tencent. I can really understand this current "fervor," as I’ve gone through similar moments myself.

A year ago, when I first tried programming agents, they probably had only a 30% success rate, but just getting a little bit right would yield a strong dopamine feedback. You could also realize that this could fundamentally change the industry, and this is their "worst moment"; the future will only get better. At that moment, I realized I could almost build anything because everything was speeding up.

Now imagine if you’re not a tech person but a small business owner suddenly discovering: “It can read my emails, manage my schedules, write Google Docs, connect to my home devices, check WhatsApp, respond to customer requests…” you would have the same epiphany as engineers have had over the past year.

During that time, I even had insomnia because this change was truly disruptive. I'm glad to bring more people from different backgrounds closer to AI.

Bloomberg: OpenAI's Codex has recently grown rapidly. What do you think about the combination of Codex and OpenClaw?

Steinberger: One core issue we currently face is: how to help users understand that a product named "programming" is actually much more than just programming.

If you look at it from a longer-term perspective, all prompts would become more powerful due to programming capabilities. AI agents are smart enough to know their shortcomings and then compensate for them by writing code.

So, does the distinction between "what is a programming tool and what is not a programming tool" still hold significance? This is also a conclusion we’ve reached internally at OpenAI. In the future, this distinction will no longer be important, so ultimately it needs to be integrated into one.

Bloomberg: What if the agent can access all your files and run continuously?

Steinberger: This is actually a question of "how to explain it to the user." Right now, you can connect almost everything in ChatGPT's application ecosystem, such as Slack, Google Docs, Notion, health data, etc. But the challenge currently is how to make users truly understand that these capabilities are already available.

Another challenge is that if you’re doing open-source projects, you can move quickly because users are more forgiving, knowing it’s a preview version and will not be used for work data. But once real work data is involved, the issues become entirely different and require more time to refine.

I look forward to participating in solving these problems.

Bloomberg: How is the progress of the OpenClaw foundation? Does OpenAI support it?

Steinberger: I try not to let OpenAI get too involved because this project needs to maintain its independence. Legal and organizational structures need a few more weeks to be finalized.

Currently, we have some great partners, such as NVIDIA, and we have been in communication with Microsoft, ByteDance has joined, and Tencent is in the process. I hope to maintain a "Swiss-style neutrality."

Our goal is to spark interest in AI among more people and truly encourage them to start thinking about problems with AI. The most critical aspect for the future is to get more people to spend more time understanding what AI can do to prepare society as a whole. This is the best way to ensure a bright future.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。