WeChat OpenClaw AI integration deepens China’s tech race

Sarah Mills | GlobalBeat

Tencent has embedded the OpenClaw AI assistant inside WeChat, the company announced on Tuesday, placing its 1.3 billion user base at the centre of China’s escalating artificial-intelligence contest.

The upgrade pushes OpenClaw into every chat window, payments screen and mini-program, giving the system access to real-time social graphs and transaction data that rivals cannot match, executives told reporters in Shenzhen.

Beijing has made domestic AI supremacy a national priority, linking future growth targets to self-developed large models and ordering state agencies to favour local providers. Tencent’s move answers similar roll-outs by ByteDance, Baidu and Alibaba, all racing to lock consumers into ecosystems that feed their algorithms fresh data each day.

“From today, every message you type can be completed, translated or turned into a service request without leaving the chat,” Tencent president Martin Lau said. “The model improves with each interaction because it sees intent, payment and follow-up in one闭环.”

Company engineers said OpenClaw now holds 130 billion parameters after continuous training on Chinese-language logs scraped from WeChat Moments, public accounts and mini-program usage. Benchmark scores released by the China Electronics Standardization Institute placed the system second only to Baidu’s Ernie 4.0 in domestic comprehension tests, though independent verification remains limited.

ByteDance reacted within hours, saying it will merge its Doubao bot more deeply into short-video app Douyin and workplace tool Feishu. Baidu shares slid 4.7 percent in Hong Kong while Tencent gained 3.2 percent, trimming the search giant’s year-to-date advance to 18 percent.

A commerce ministry notice circulated last week told local governments to “prioritise certified domestic models when procuring smart-city software,” a provision analysts peg at $8 billion in annual contracts. Government clients already account for 38 percent of Baidu’s AI cloud revenue, the company disclosed in May.

Privacy groups warned that embedding algorithms inside chats that include location, medical and financial data raises fresh risks. “WeChat OpenClaw AI retains inputs by default to refine replies, creating a trove that state bodies can requisition under cybersecurity rules,” Peking University scholar Wang Lina said. Tencent replied that users may switch to “incognito mode” to prevent retention, though doing so disables personalised features.

The integration follows new rules that cap foreign shareholding in Chinese AI ventures at 49 percent and require security reviews for models trained on more than 10 billion parameters. Washington has tightened curbs on advanced chip exports, cutting Tencent and peers off from Nvidia’s latest GPUs and forcing them to train with lower-grade substitutes or smuggled stock.

Background

China launched its “New Generation AI Development Plan” in 2017, setting a target to match the West by 2020 and lead globally by 2030. The blueprint channelled grants, subsidies and tax breaks worth $56 billion into domestic labs, according to ministry disclosures compiled by brokerage Jefferies.

ChatGPT’s November 2022 release startled Beijing regulators, who saw U.S. models pulling ahead while China tightened speech controls that limit training corpora. Authorities responded by green-lighting over 200 local large models, yet only 40 have received public-facing permits, forcing others to market enterprise tools.

What’s Next

Ant Group and JD Technology plan to unveil updated bots next month, people familiar with the matter said, intensifying competition for consumer data ahead of the annual Singles’ Day shopping festival that drives half of China’s yearly e-commerce volume.

Regulators will watch whether deeper AI integration inside super-apps accelerates misinformation or fraud. The cyber-affairs watchdog has asked Tencent, ByteDance and Alibaba for weekly metrics on hallucination rates and complaint volumes, an official told Reuters, signalling tougher oversight if incidents spike.