Since the beginning of 2026, an open-source artificial intelligence agent called OpenClaw has rapidly gained popularity in the global technology community. AI tools that can perform tasks in the real world and take actions on behalf of users have begun to enter the public eye, attracting widespread attention. Some observers believe that AI agents capable of autonomously executing tasks demonstrate strong technological potential and may drive AI from "conversational assistants" to "action assistants", ushering in a new era of personal digital assistants. At the same time, American media and cybersecurity experts have frequently issued warnings, pointing out that there are still risks in terms of security, reliability, and other aspects, and their high access to user devices has also raised concerns about data security and privacy protection. OpenClaw, developed by Austrian software engineer Peter Steinberg, is an open-source AI intelligent agent software that has attracted much attention for its technological potential. Unlike traditional chatbots, this intelligent agent can autonomously perform complex tasks such as file management, email sending and receiving, and data processing on the user's local computer by integrating communication software and large language models. In addition, users can also train and expand their abilities by installing skill pack code. Public data shows that OpenClaw received over 300000 "stars" in just over two months after its launch on the open-source platform GitHub. This indicator is often seen as an important reference for the popularity of open source projects, indicating a high level of attention from the developer community. At the same time, OpenClaw has also spawned some creative projects, such as Moltbook, a network platform that allows millions of AI agents to autonomously socialize. The American technology media generally believes that this "mobile AI agent" represents a new direction in the development of AI, which may change the way humans interact with computers. By connecting email, scheduling, and various software systems, AI can automatically complete a series of complex tasks, thereby improving the work efficiency of individuals and businesses. Industry insiders point out that if the relevant technologies gradually mature, such AI intelligent agents are expected to be widely applied in fields such as office automation, digital assistants, and enterprise process management in the future. The rapid popularity of OpenClaw has also raised concerns among security experts and the media regarding its security and reliability. American media believes that such AI agents require high system privileges to perform tasks such as accessing files, running programs, or connecting user accounts. Improper configuration may lead to data breaches or even malicious exploitation. Network security experts say that such systems are vulnerable to a security technique attack against large language models called "prompt injection attack". Attackers may induce AI to perform unauthorized operations by hiding instructions, thereby obtaining sensitive data or manipulating the system. OpenClaw developer Steinberg clearly stated in a blog post on his company's official website that "prompt injection attacks" are still an unsolved problem in the industry. According to researchers cited on the Forbes magazine website, the reason why AI agents are causing security concerns is because they possess three high-risk features simultaneously: automated systems capable of autonomously executing tasks, information sources that may hide malicious instructions, and the ability to access user devices with high privileges. According to TechRadar, a technology media platform, security researchers have discovered a major security vulnerability called "ClawJacked" in the core system of OpenClaw. Attackers may take over AI agents through malicious web pages, thereby obtaining device permissions and accessing system data. The threshold for using OpenClaw is still high. In addition to security issues, many users also express that the threshold for using OpenClaw is high. A user posted on technology forums such as the Reddit website in the United States, stating that the process of installing OpenClaw was a "nightmare" with various compatibility issues and errors that ultimately led people to give up using it. Some users also expressed that OpenClaw has a complex operating environment, insufficient stability, and a user experience that is "too expensive, too slow, and not reliable enough". TechCrunch, a US technology news media, reported that OpenClaw security configuration and operation require high technical skills, and currently the tool is more suitable for technical personnel or developers to use, and is not suitable for ordinary public users. Steinberg stated in a media interview that OpenClaw requires careful configuration to ensure security and is currently not suitable for non-technical users. Overall, the American media and technology community generally believe that OpenClaw represents an important direction for AI development - from "conversational AI" to "action AI". However, this type of technology is still in its early stages of development, and its security, stability, and regulatory framework still need further improvement. Network security experts point out that as AI agents gradually gain access to real-world systems and data, their potential risks are also expanding simultaneously. How to improve security governance and regulatory systems while promoting technological innovation has become an important issue that artificial intelligence must face as it moves towards wider applications. (New Society)