These papers are currently in draft status and should not be cited
James, Andrew. April 2026. AI in Financial Markets: A framework for AI induced market endogeneity.
Abstract
The adoption of artificial intelligence in finance is widely assumed to enhance market efficiency by reducing human bias and improving information processing. This paper challenged this assumption and argued that artificial intelligence systems in financial markets do not enhance efficiency, instead they replicate and enhance historically embedded inefficiencies. Challenging the assumption that data-driven trading leads to more rational markets, by showing how financial data encodes behavioural bias and structural constraints that persist under limits to arbitrage. Drawing on behavioural finance, limits to arbitrage and reflexivity, the paper will propose a conceptual framework of AI-induced market endogeneity, in which AI models become endogenous to price formation rather than an external optimiser. The paper also suggests a regulatory approach for AI models in financial markets including a discussion on model governance, transparency and accountability.
James, Andrew. April 2026. A contractualist argument for rejecting AI in financial markets.
Abstract
Artificial intelligence adoption in financial markets has been growing since the early 2000s, however the interest in AI has grown significantly in the last 10 years with an eye to improved market efficiency from rational AI agents. This paper argues that the inherent human biases that exist within financial markets mean true rational AI agents cannot be trained on financial market data. Due to the reflexivity of financial markets increased adoption of AI in financial markets means as adoption increases AI agents will move from external optimisers to endogenous components of financial markets. In financialised commodities markets the risk of this endogeneity may lead to increased financialisation. Given the potential risk imposed by AI endogeneity in financial markets this paper argues that in line with contractualism, including ex post contractualism, AI should not be adopted in financial markets. This paper will present the argument that the risk of AI induced endogeneity will sit disproportionately with those in lower socioeconomic categories as the magnitude of harm is greater for them than those of higher socioeconomic backgrounds. As such AI should not be introduced into financial markets unless strong governance and policy are put in place to ensure that harm reduction is ensured.
Available here: https://rpc.cfainstitute.org/blogs/enterprising-investor/2026/vix-policy-uncertainty-risk-signal