Zhou Hongyi first proposed "Skill Security": AI intelligent entities enter the "hands-on" era, and safety is no small matter

On the afternoon of March 12, 360 Group held a "Lobster Security Media Exchange Conference" in Beijing, where the company's founder Zhou Hongyi interpreted the recently widely discussed AI intelligent entity OpenClaw (Lobster) and focused on responding to external concerns about "Lobster Security." Image: Scene of the "Lobster Security Media Exchange Conference" At the conference, Zhou Hongyi stated that "Lobster is a good thing, definitely not a so-called virus." He believes that it is normal for new technologies to spark controversy, and "not developing or progressing is the biggest security risk." In his view, traditional intelligent entities are more like tools with preset capabilities, while "Lobster" resembles an evolved entity with autonomous capabilities. In the future, users may not just be "raising an intelligent entity," but managing a large number of AI skills. If the popularization of "Lobster" can drive the implementation of intelligent entities in government agencies and enterprises, and help the public accept the concept of intelligent entities, it will be a significant contribution to the development of China's AI industry. He also pointed out that the AI industry is entering a dual-line evolution stage of "large models + intelligent entities," and "Lobster" is confirming this trend: large models continuously enhance cognitive abilities, while intelligent entities continuously strengthen execution capabilities. As AI evolves from "being able to answer questions" to "being able to take action," the application of intelligent entities is also beginning to extend to more practical scenarios. Zhou Hongyi further stated that "Lobster" represents a new paradigm of intelligent entities that can combine the capabilities of large models with business operations, promoting application implementation through a "small incision, large depth" approach. This "seeing the big from the small" method is particularly suitable for the intelligentization and informatization implementation of Chinese enterprises. Currently, an ordinary computer can run intelligent entities, accelerating "one-person company" (OPC) style entrepreneurship. At the same time, as the capabilities of intelligent entities continue to enhance, related security issues are gradually receiving attention. Zhou Hongyi pointed out that Lobster, as a new entity, should not be simply denied due to potential risks, but rather, safety boundaries should be gradually established during its development. He reminded ordinary users to pay special attention to account and fund security when using intelligent entities and not to disclose sensitive information such as passwords and codes to intelligent entities. It is worth mentioning that the day before this media exchange conference, 360 released the "OpenClaw Security Deployment and Practice Guide," which systematically analyzed the potential security risks of OpenClaw and proposed corresponding solutions. Zhou Hongyi stated that many people are currently discussing "raising lobsters," but there are still questions about "how to safely raise lobsters." As a company that has long been deeply involved in the field of cybersecurity, 360 has the responsibility to fill the security gap during the rapid development of new technologies, laying a solid security foundation for the development of new technologies such as intelligent entities