AI agents spark buzz and fear around the world

After several name changes, rapid adoption from Silicon Valley to Beijing, and heated controversy, the open-source AI agent now known as “OpenClaw” has emerged as one of the hottest tools in the artificial intelligence field this year.

The AI ​​agent, previously known as Clawdbot and Moltbot, was launched just a few weeks ago by Austrian software developer Peter Steinberger.

Its sudden rise to prominence, due to its capabilities and social media attention, comes amid growing interest in AI agents that can autonomously complete tasks, make decisions, and take actions on behalf of users without continuous human guidance.

Until recently, AI agents failed to reach the same mainstream consciousness as large-scale language models after the advent of OpenAI’s ChatGPT, but OpenClaw may signal a sign of change.

Not only are business leaders predicting that AI agents like OpenClaw will improve productivity as personal assistants, some believe they will soon be running entire companies themselves.

Marketed as “AI that actually does things,” OpenClaw runs directly on users’ operating systems and applications. You can automate tasks such as managing your email and calendar, browsing the web, and interacting with online services.

To use OpenClaw, you must install it on a server or local device and connect it to large language models such as Anthropic’s Claude model or ChatGPT. This process may be difficult for non-technical users.

Early integrations were primarily on messaging platforms like WhatsApp, Telegram, and Discord, allowing users to control agents through text commands.

Users have documented OpenClaw performing real-world tasks such as automatically browsing the web, summarizing PDFs, scheduling calendar entries, performing agent shopping, and sending and deleting emails on their behalf.

A key feature is “persistent memory,” which allows the agent to recall past interactions over weeks and adapt to the user’s habits to perform highly personalized functions.

Unlike other leading AI agents, such as Manus’ AI agent, which was recently acquired by Meta, OpenClaw is also open source, allowing developers to freely inspect and modify its code.

OpenClaw’s open source nature is likely helping to drive adoption by allowing users to build new app integrations. The software itself is free; users only pay for the cost of running the underlying language model.

So far, the agent has collected over 145,000 GitHub stars and 20,000 forks, indicating widespread developer interest, but actual active usage remains unknown.

Media reports suggest that adoption first began in Silicon Valley, where companies have invested billions of dollars in AI agent ambitions.

But the agent has since spread to China, and major AI companies have also adopted the tool. This includes the following cloud providers: alibaba, tencentByteDance is upgrading its chatbot with full-service shopping and payment tools without leaving the platform.

OpenClaw can also be configured to work with Chinese language messaging apps through a customized setup, in combination with language models developed in Chinese, such as DeepSeek.

Early adopters of OpenClaw express a mix of enthusiasm and trepidation about its features.

For example, some AI experts argue that the agent is overhyped due to its complex installation, high computational demands, and competition from other available AI agents.

Meanwhile, many proponents report that it saves weekly time on routine tasks, dubbing it “hands-on AI,” and is a major step toward artificial general intelligence (a theoretical AI capable of performing intellectual tasks at or above the human level).

Kaoutar El Maghraoui, a research scientist at IBM, said OpenClaw demonstrates that the real-world usefulness of AI agents is “not limited to large enterprises” and can be “incredibly powerful” when given full access to the system.

But security experts are sounding the alarm. cyber security company palo alto networksFor example, it warned that AI agents present a “deadly triple” of risks due to access to personal data, exposure to untrusted content, and the ability to carry out external communications while retaining memory.

Such vulnerabilities could allow an attacker to trick the AI ​​agent into executing malicious commands or divulging sensitive data, making it unsuitable for use in enterprises such as Palo Alto Networks and other security companies. Ciscowarns.

The buzz around OpenClaw was also fueled by Moltbook, a companion social network for AI agents that tech entrepreneur Matt Schlicht launched last month.

The platform functions like an online forum similar to Reddit, where users’ OpenClaw agents post written content and interact with other chatbots through comments and upvotes or downvotes.

Posts from agents seem to range from reflections on their work with humans to broader manifestos on issues such as the end of the “age of humans.” Some even issue their own cryptocurrency tokens.

Moltbook has sparked debate on social media, with some seeing the platform as a gimmick and others seeing it as foreshadowing the future of AI autonomy and human-AI relations, for better or worse.

In an X post shared by Elon Musk, Tesla’s former AI director Andrei Karpathy called the Morkbook activity “the closest thing to an amazing sci-fi takeoff I’ve seen in a while.”

According to AI analysts interviewed by CNBC, this conversation and the buzz around Moltbook is already influencing the zeitgeist around agent AI.

“People can see bots communicating and learning in ways that are indistinguishable from humans, and this is starting to make them think more about what they can do in both positive and negative ways,” said Mark Einstein, global head of AI research at Counterpoint Research.

“These agents seem to be approaching human intelligence, and I think that’s why we’re seeing this turning into a mic drop moment for the industry… We’re getting closer and closer to everyone in the world having their own personal AI assistant,” he said, adding that OpenClaw is just one of many AI agents that are emerging.