Summary of Key Points
WeChat's new Mini AI assistant offers basic functions such as summarization, Q&A, and content generation, which are not significantly different from other AI services like DouBao and Kimi. However, by leveraging the comprehensive data on WeChat's 1.4 billion users—including their social interactions, consumption habits, content preferences, and work activities—this assistant has the potential to become the most user-aware AI platform. In practice, when accessed through the main entry point (by swiping left), Mini feels like a "strange AI" that requires users to actively provide input. However, in contexts such as reading articles, watching videos, or participating in group chats, Mini functions smoothly without the need for users to switch apps or provide additional context. The paradox here is that despite possessing vast amounts of data (a "gold mine"), super platforms like WeChat and Apple are hesitant to delve deeper due to concerns about user trust and privacy. Data is crucial for making AI useful but also forms the foundation of user trust, creating a dilemma for these platforms.
1. The Duality of Mini: An Unassuming Main Entry Point, an Attentive Assistant in Specific Contexts
The experience with Mini is quite divided:
- Main Entry Point (Swipe Left): A Disappointing “Strange AI”
Swiping left to access Mini opens a blank dialogue box where users must type their requests (e.g., "Summarize this article for me"), just like using DouBao or Kimi. On the first day, I could count the number of times I actively used this feature with one hand—after all, the act of deliberately swiping left made me question its relevance to my current needs.
- Embedded in Context: A User-Friendly Assistant
But within specific contexts, Mini becomes much more responsive. For example, while reading a public account post, you can click on highlighted text to ask Mini for an explanation without copying and pasting; at the bottom of an article, clicking “Ask Mini” allows you to get a summary along with further details; when watching a video, you can hand over the video to Mini, and it will extract key information and main points; in group chats, asking Mini to summarize the discussion is straightforward. In these scenarios, Mini understands what you are viewing and discussing without the need for additional context, providing a seamless experience.
2. WeChat’s “Data Gold Mine”: A Treasure trove, Yet Hesitant to Be Exploited
WeChat holds one of the most extensive user datasets in China:
- How Comprehensive is the Data? It includes your chat history, contact information (who you frequently interact with), payment records (where and how much you spent), content preferences (followed public accounts, watched videos), and work-related data (group chat files, activity logs). This collection creates a "digital profile" of you that even your partner might not know in such detail.
- Why Hesitate to Exploit It? WeChat’s core value is user trust. 1.4 billion users trust it with their private information because it has consistently protected their data over the past 14 years. For instance, while checking bills, Mini can provide a weekly expenditure summary, which is convenient, but it raises concerns: "Does it know about all my purchases?" Every additional data permission makes Mini more useful, yet it also weakens user trust.
3. The AI Dilemma for Super Platforms: Useful AI Requires Data, Yet Data Is a Threat
This issue is not unique to WeChat; Apple faces the same challenge:
- Apple’s Siri AI: Delayed Release Due to Privacy Concerns, Resulting in a $250 Million Loss**
Apple has been promoting its Siri AI for two years, but core features (reading texts, browsing photos, cross-app operations) are still not fully available. In 2024, the iPhone 16 ads promised an enhanced Siri AI, which led to a class-action lawsuit and a $250 million settlement. The reason is simple: these features require access to user privacy data, and Apple is reluctant to grant such permissions.
- The European Union Market: Prefer Not to Launch Features Over Sharing Permissions**
This year, the EU’s Digital Markets Act requires companies like Apple to share underlying device permissions with third-party AI assistants. Apple decided not to launch Siri AI in the EU, preferring to lose the market rather than compromise user privacy.
- The Twisted Nature of Data: Data is a double-edged sword. Super platforms need data to make AI more user-friendly, but using it can erode trust, which is their most valuable asset. This creates a dilemma: more data equals better AI, but more data also risks damaging trust.
4. Different Approaches Based on Data Availability
AI providers with varying amounts of data adopt different strategies:
- Independent AI (ChatGPT, Claude): Limited Data, Less Restrictive
They start with no user data, so they can be more experimental, but users must provide background information (attach files, describe tasks) to train the AI.
- Mobile Device Manufacturers’ AI: Partial Access to Data, A Moderate Approach**
AI assistants from companies like Xiaomi and Huawei have access to some user data, allowing them to understand users better than independent AI services. However, their data volume is still smaller compared to WeChat and Apple, so they are more cautious (e.g., helping organize photos).
- Super Platforms (WeChat, Apple): Comprehensive Data, Extreme Caution
With the most data, these platforms are the most hesitant, as users store their most private information here. Any mistake could result in significant losses. Therefore, Mini can only use data when users initiate requests, for read-only purposes, and within a limited time frame (e.g., summarizing recent group chat history).
5. WeChat’s Restraint: A Strategic Choice, Not Technological Limitation
Many believe Mini lacks intelligence, but it’s not due to technical limitations:
- What Can Be Done Technically? Functions like sending bulk messages or automatically responding to all chat records are possible, but WeChat avoids them to prevent spamming friends and maintain a healthy social environment.
- Why the Restraint? WeChat’s strength lies in user trust, not AI capabilities. While AI features can be improved over time, losing trust is irreversible. Mini’s “ordinary” performance reflects WeChat’s cautious approach in testing user acceptance limits without taking risks.
In Conclusion
In the AI era, super platforms’ access to vast data is both an advantage and a constraint. To make AI more user-friendly, they need to collect data, but this can lead to trust erosion. WeChat’s relatively basic Mini AI reflects its careful balancing of functionality and security. This balance is a critical challenge for all super platforms.