The Pulse
Ant Group Launches LingGuang 2.0 as a Persistent Personal Agent
Ant Group launched LingGuang 2.0 on October 27, adding cross-session memory and tasks that can continue on cloud instances after a conversation ends. The assistant is rolling out in China, where users can authorize it to monitor Alipay acti
AI.info Team · Written with AI · fact-checked ·
Ant Group launched LingGuang 2.0 on October 27 with a change in job description: its AI assistant is designed to keep working after a user leaves the conversation. The company says the upgrade adds cross-session memory, long-term task scheduling and tool use, allowing the assistant to monitor changing information and follow up when something needs attention. The service is now rolling out in China on iOS, Android and LingGuang.com.
Tasks that outlast a chat
LingGuang 2.0 runs tasks on independent cloud instances, according to Ant, rather than limiting work to a single exchange with the user. With explicit authorization, it can draw on connected information and services, begin analysis or monitoring, and alert a user when a change matters. Ant presents that continuity as the difference between completing one request and managing an ongoing task.
The company’s examples include reviewing Alipay spending records to flag potentially forgotten auto-renewing subscriptions, analyzing spending patterns and suggesting savings. After a flight or hotel booking, the assistant can monitor price changes and notify the user if a lower fare or rate appears. Ant also names reminders tied to pet care, family birthdays and vehicle inspections.
Alipay, work apps and smart-home tools
The assistant combines services in the Alipay ecosystem with connections to email, DingTalk, Lark and Mi Home, Ant says. It can also call tools through MCP—the Model Context Protocol—and command-line interfaces in open environments. That mix gives the product a route from personal records and travel services to workflows spanning other apps and smart-home devices.
The range of connections also makes permission management central to the product. Ant says users must authorize access for LingGuang to act on connected information, and can review, change or revoke those permissions. The company says its APASS security service separates authorization keys from the agent and holds them in escrow, so the agent can perform tasks without directly accessing credentials.
Ant’s claims meet the practical test
Those safeguards and examples describe Ant’s intended design; the launch announcement does not provide independent testing of how reliably the assistant handles changing plans, catches relevant updates or respects user-set boundaries in real use. Nor does it specify how often the system checks connected services or what users will see when it proposes an action. Those details matter when a tool is expected to act across multiple accounts and keep doing so beyond the active chat.
LingGuang first launched in November 2025, and Ant says it passed 2 million downloads in its first six days. The company had previously added prompt-based Flash App generation and world-model features; version 2.0 shifts the emphasis to following tasks over time. For now, the rollout is limited to China, and Ant’s announcement does not give a date for availability elsewhere.
Permission is the dividing line
LingGuang’s most consequential promise is also its most demanding: the assistant should identify a need and keep pursuing it without waiting for another prompt, but only within permissions the user has granted. Whether that balance feels helpful will depend on the accuracy of its alerts, the clarity of its controls and the services it can access. Users can begin trying the China rollout on iOS, Android or LingGuang.com.