China forum urges tighter AI risk controls in digital finance boom

Beijing is effectively signalling that large-scale deployment of AI agents in finance is now a fait accompli – and that the policy focus is shifting from ‘if’ to ‘how safely’. When former Bank of China leadership frames agent-based systems as already active in quant trading, risk, and advisory, and regulators talk about hallucinations and model opacity, you are seeing a major financial system treating frontier AI as critical infrastructure. That matters for the race to AGI because it locks in a demand flywheel: as financial institutions commit to agentic workflows, they create a recurring need for more capable, more controllable models.

At the same time, China is experimenting with embedding fairness, explainability and compliance constraints deeper into model training for financial use. If those requirements are enforced, they could nudge Chinese labs toward architectures that are more interpretable but potentially marginally less aggressive than unconstrained frontier systems. The global impact is a two‑speed pattern: heavily regulated, highly capitalised sectors like banking will demand safer agents, while less regulated verticals keep pushing raw capability. That mix probably doesn’t slow AGI much, but it does shape where the sharpest edges show up first.