At the 2026 AI DingTalk 2.0 new product launch event today, DingTalk's founder and CEO, Chen Hang, introduced Wukong, an AI-native work platform.
Regarding Wukong’s security controls, DingTalk CTO Zhu Hong explained that the team adopted an integrated "software-hardware" design approach from the outset and established a six-tiered security framework: The first layer involves basic security rules, including control over computer system operation permissions and fundamental access rules for personal computers. The second layer is unified identity authentication; using Wukong requires enterprise-level account authorization, and Wukong’s permissions are less than or equal to those of employee accounts, with all operations conducted within authorized limits. The third layer provides enterprise-specific capabilities, allowing companies to develop and deploy internal system skills to Wukong, avoiding security risks from complex manual operations. The fourth layer includes network proxy and monitoring; Wukong can access external networks but is subject to layered network policies to ensure every network access is traceable. The fifth layer is an AI-native file system, custom-built for Wukong to ensure all data changes are traceable and auditable. The sixth layer covers operational environment control, giving enterprises full authority over Wukong’s runtime environment to ensure compliance with enterprise-grade security standards.
When asked how Wukong differs from other AI Agent products like OpenClaw, Zhu Hong noted that while there are over a dozen similar "Claw" products on the market, two fundamental issues remain unresolved: usability and security. He stated that beyond engineers, general users often struggle to operate such tools correctly, leading to services like on-site installation for a fee. Wukong, by contrast, lowers the usage barrier significantly, enabling out-of-the-box usability similar to installing DingTalk. Additionally, security was a core consideration in Wukong’s design from the beginning.
Zhu Hong further revealed that Wukong’s development was not instantaneous but involved nearly a year of incubation. The idea emerged in May of last year with a small exploratory team; by August, it was formally established as a project; and by December, a product was delivered for user testing. The team has since expanded from two initial members to 20-30 people, with more resources expected to be allocated going forward.
On the business model, Zhu Hong outlined a platform-oriented vision, expressing hope that Wukong could become a "Taobao" for AI in the enterprise sector, attracting traditional software companies to integrate and monetize computing power through capability sharing. He emphasized a goal of shifting toward a pay-per-use and pay-for-performance model, which will require industry-wide participation to benefit the entire ecosystem.
Zhu Hong also disclosed that Alibaba employees will be the first co-creators of Wukong, providing a reference for the industry. This means that before its market release, Wukong will undergo large-scale testing among tens of thousands of Alibaba staff, refining the product while accumulating enterprise application scenarios.
Addressing concerns about AI replacing human labor, Zhu Hong offered an optimistic outlook, noting that historically, industrial revolutions have not led to mass unemployment but instead introduced more convenient tools. He emphasized that the team’s philosophy is to use AI to enhance human productivity, and that AI’s emergence also creates many new types of jobs.
Comments