Executive Summary:
Artificial intelligence is rapidly becoming embedded in investment research, due diligence, manager selection, portfolio construction, and risk management. Much of the industry discussion has focused on the productivity gains AI can deliver by processing vast amounts of information more quickly than humans. This paper challenges the assumption that reading more information necessarily leads to better decisions.
The author introduces the concept of technological epistemic risk: the risk that the deployment of AI changes the information environment it is designed to analyze. Once firms understand which signals, metrics, and disclosures AI systems reward, they may begin optimizing their communications for the machine rather than for human investors. In this setting, AI may produce a paradoxical outcome: investors gain access to more data and more analysis, yet acquire less genuine knowledge about underlying economic realities.
Using examples from private markets, due diligence, disclosure practices, and investment decision-making, the paper demonstrates how AI-driven information extraction can create incentives for strategic behavior and disclosure manipulation. The central insight is that historical predictive power alone is no longer sufficient for evaluating information quality. Investors must also assess how costly it is for a signal to be manipulated once market participants know that AI systems are using it.
For senior financial professionals, boards, and regulators, the paper offers an important warning: the greatest risk of AI may not be that machines misread information, but that their widespread adoption changes the incentives of those producing it. As AI becomes increasingly integrated into investment processes, governance, verification, transparency, and human judgment will remain essential safeguards against a future in which investors have more information, but less insight.
Three Main Takeaways
- AI can create technological epistemic risk by changing the information environment it is designed to analyze.
- As firms learn how AI systems extract and evaluate information, they may increasingly tailor disclosures to AI systems, potentially leaving investors with more data but less genuine knowledge.
- Investors should evaluate information not only on its predictive power, but also on how difficult it is for market participants to manipulate or “game” the signals that AI systems rely upon.
Special Issue on Emerging Risks by GRI in partnership with the Journal of Portfolio Management