Zhou Hongyi discusses AI risks: Lobster-type intelligent agents still pose security risks, "secure intelligent agents" become the key solution

On April 19, the China Internet Association held the 2026 China Internet Development Symposium in Beijing. At the meeting, academicians, experts, and corporate representatives exchanged views on topics such as artificial intelligence, intelligent agent innovation, security governance, and the synergy between computing power and energy. Zhou Hongyi, the founder of 360 Group, attended the meeting as a representative and delivered a speech. Image: Zhou Hongyi, founder of 360 Group, speaking at the China Internet Development Symposium In his speech, Zhou Hongyi stated that artificial intelligence is transitioning from the "large model stage" to the "intelligent agent stage," with the industry's focus shifting from "being able to converse" to "being able to execute." Intelligent agents, represented by "lobsters," now possess the ability to call tools and autonomously complete tasks, beginning to enter actual production processes. He pointed out that intelligent agents are no longer just tools but "digital experts" capable of undertaking professional work, which will reshape the way humans and machines collaborate. He believes that with the popularization of intelligent agents, companies will gradually form a hybrid team structure of "carbon-based employees + silicon-based employees," where employees need to have the ability to manage and collaborate with intelligent agents. At the same time, intelligent agents will lower the barriers to entrepreneurship and promote the development of new models such as "one-person companies." At the industrial level, Zhou Hongyi noted that intelligent agents will drive the reconstruction of internet infrastructure, with future software forms becoming more modular, and even generated by intelligent agents. He predicts that in the next three to five years, China is expected to form a scale of intelligent agents worth hundreds of billions, which will become the main force of the internet and give rise to a new intelligent economic ecosystem. Regarding computing power and energy, Zhou Hongyi stated that the application of intelligent agents will bring far higher computing demands than traditional AI, with token consumption potentially increasing dozens or even hundreds of times, thereby driving the collaborative development of computing power, chips, and electricity industries. He also emphasized that China's advantages in energy infrastructure will become an important support in the era of intelligent agents. In terms of security, Zhou Hongyi pointed out that intelligent agents still face uncertainties and new types of attack risks in enterprise applications, making security a key factor for implementation. He proposed "using models to govern models," advocating for the use of security intelligent agents to counter risks and promote the development of artificial intelligence under controllable conditions. It is reported that the China Internet Development Symposium has been held continuously for 22 sessions since its establishment in 2004, serving as an important platform for gathering industry consensus and assessing industry trends. This meeting focused on the "Artificial Intelligence +" initiative, emphasizing driving industrial upgrades through technological innovation and ensuring development through security. This aligns with Zhou Hongyi's judgment that "intelligent agents will reshape productivity, and computing power and security will become key supports." Zhou Hongyi stated that in the future, 360 will continue to adhere to the "Security + AI" strategy, actively participate in the "Artificial Intelligence +" initiative under the guidance of regulatory authorities, and promote the coordinated development of technological innovation and security governance to support the high-quality development of the internet industry