Inside China's bank AI adoption: on-site bans on Claude Code, weak local 30B models, humans still coding
aigclink · x · 2026-08-30
The author talked with a friend with 10 years in fintech, revealing the real struggles of AI adoption in China's banking sector:
- Fintech vendors are struggling: industry crackdowns, pressure from bank tech departments, and payment cycles stretching to 5 years have hurt former internet-finance stars; the high-end body-shopping model is being taken over by bank IT + AI.
- Strict on-site rules: bank projects require working on-site under camera surveillance, with cloud tools like Codex and Claude Code banned—most coding is still done by hand.
- Localization is the biggest blocker: banks demand local deployment but budgets are thin, forcing weak 30B/100B models that can't solve real problems. The only viable path is "financial FDE" work—big brands win contracts, FDEs deliver, essentially body-shopping with a trendy title.
- Securities is closed to outsiders: Tonghuashun and Eastmoney already have their own AI assistants, leaving little room for external players.
More from Companies & People
- QuiverAI provides first creator access to Arrow 2.0 model — stuffyokodraws · 2026-08-30
- Finance+AI on the Ground: Only Three Directions Still Hold Opportunity — aigclink · 2026-08-30
- SF engineer criticizes city's shift to arbitrage, launches Engineering Society — KadriJibraan · 2026-08-30
- 21-year-old founder rejects CMU, drone startup raises $37M — sanjaykalra · 2026-08-30
- OpenAI employee states company incentives run counter to math community goals — ElliotGlazer · 2026-08-30
- William Shatner discusses AI and knowledge in new interview — GregCook2011 · 2026-08-30