Efficiency and Fragility as Joint Products: AI, Governance, and Financial Markets in the US and India

Artificial intelligence now makes most of the decisions in financial markets. Algorithms execute the majority of trading in large equity markets; JPMorgan has estimated that discretionary human traders account for only about a tenth of US equity trading volume, with quantitative and passive strategies making up the rest (Kolanovic, 2017), and the same tools price assets, construct portfolios, score borrowers, detect fraud, and interpret textual information such as news and sentiment. This development is conventionally framed as a trade-off, in which AI makes markets faster, cheaper, and more liquid at the cost of making them less stable and less fair. This paper argues that the trade-off framing is incomplete. Efficiency and risk are not two separate effects to be set against each other. They arise from a single change: the replacement of rules written by people with rules learned by machines from data, a change that produces the gains and the dangers together. Speed, scale, and pattern recognition narrow spreads and widen credit access, and the same three properties produce flash crashes, herding, opacity, and bias in lending. Because both sides share a source, the net effect of AI on markets is not determined by the technology. It is set by the rules built around it, which makes regulation the deciding variable. The argument is developed through formal models, secondary market data, and documented cases, using the United States as the mature reference case, where the long data series and the studied failures reside, and India as the live test, a younger and far more retail-heavy market in which adoption has outrun oversight. The central finding is that whether AI helps or harms markets is a question of governance rather than of technology.