In 2025, China’s CSRC launched “SkyNet 2025,” a real-time surveillance system leveraging federated learning to detect insider trading—without breaching data privacy.
How It Works: Privacy-Preserving Policing
Fifteen major brokerages contribute encrypted model gradients rather than raw client data. Using a technique called gradient anomaly detection, SkyNet flags accounts whose trading model deviates more than 3 standard deviations from the norm. It’s the AI equivalent of a lie detector for trading intent.
Result: Within one month, the system froze ¥4.3B in “shadow-paired trades”—a covert tactic where two linked accounts simulate market noise while extracting illegal alpha.
The Rise of Proactive Regulation
Traditional financial oversight has been reactive, with penalties issued months after damage. Now, compliance AI enters the arena mid-game, freezing funds and issuing alerts before the bell even rings.
A New Arms Race: Compliance AI vs Rogue AI
The implication is profound:
Markets are no longer shaped by traders alone but by algorithms enforcing ethics in real time. The turning point will be when compliance AI outperforms adversarial AI, pushing markets toward a state of:
“Davis Double Purity” – a nod to the “Davis Double Kill” in investing, where shrinking valuation and profits accelerate decline. Here, however, it’s bad actors and alpha that evaporate together.