AI Disclosures as Signal or Shield: Risk, Investment, and the Informativeness of 10-K Filings

Sammanfattning

This thesis examines whether section-specific artificial intelligence (AI) disclosure in U.S. 10-K filings is associated with filing-window market reactions and future operating performance. Using 2017-2024 filings for firms in a retrospective early-2026 S&P 500 constituent snapshot, the study links SEC EDGAR text with Capital IQ accounting data, CRSP returns, and CRSP/Compustat identifiers. AI disclosure is measured separately in strategy-oriented sections, risk factors, and financial statement footnotes. The results are mixed. Filing-window tests do not show a consistently positive market reaction to AI disclosure. By contrast, strategy-oriented AI disclosure is significantly associated with one-year-ahead operating performance, especially operating income to revenue, while risk-oriented and accounting-note AI disclosure shows weaker and less stable associations. Strict observed R&D backing does not strengthen the strategy-disclosure relation. Overall, the findings suggest that AI disclosure is most informative when interpreted by section location, timing, and outcome type. The results are conditional associations rather than causal effects, and they point to strategy-oriented AI disclosure as more closely related to future operating outcomes than to immediate filing-date market reactions.