JD.com, Meituan tighten internal use of external AI tools as firms push in‑house models

It has been reported that JD.com (京东) has moved to fully block employee access to a range of external AI services, and Meituan (美团) is restricting internal recommendations of Alibaba Cloud’s Qwen model (阿里云 Qwen). The measures, reportedly implemented at the end of March for JD and rolling out at Meituan this month, mark a clear corporate shift toward favouring self‑developed models over third‑party offerings. Which platforms are affected? The reported JD blocklist includes domestic models such as 豆包 and 千问 as well as international offerings like Google’s Gemini, OpenAI’s ChatGPT, xAI’s Grok and others.

What changed inside the companies

According to the report, JD employees who try to open blocked external AI sites are redirected to an interception page that links to JD’s own large model and provides a formal application route for external AI use. It has been reported that Meituan has adjusted its recommendation policies so business teams are no longer steered toward using Alibaba Cloud’s Qwen by default; teams that still need Qwen must submit a justification to a manager (generally a x3 level) for approval. Reportedly, some models such as 豆包 currently do not require the same approval, while Meituan recommends its in‑house LongCat (龙猫) model for internal use.

Broader product push and geopolitical backdrop

These moves come as JD has been publicly consolidating its AI strategy — it has been reported that JD rebranded its large‑model family to JoyAI (覆盖 3B 到 750B 参数) at the World Artificial Intelligence Conference and is expanding multimodal capabilities — and rolled out AI assistants for frontline staff. Corporates’ preference for internal stacks should also be read against tighter regulatory scrutiny in China over data security, and a global landscape shaped by export controls and geopolitical tension that encourages tech self‑reliance. For enterprises and developers, the question is simple: will internalization accelerate Chinese firms’ divergence from Western AI ecosystems, or will demand for interoperability force new neutral pathways?