Uppsats

Artificial Intelligence Risk: Evidence from LLM-Assessed 10-K Filings, 2020-2025

Kandidat-uppsats

Handelshögskolan i Stockholm/Institutionen för finansiell ekonomi

Publicerad: 2026

Språk: Engelska

Sammanfattning

We construct a firm-level AI risk exposure index from mandatory 10-K Item 1A disclosures for S&P 500 firms over fiscal years 2020-2025. Three large language models score each firm's risk factor text across four dimensions: displacement, operational, regulatory, and cybersecurity risk. We test whether AI risk exposure predicts differential equity returns using quartile portfolio sorts, Fama-MacBeth cross-sectional regressions, bivariate sorts, and panel regressions on high AI attention days identified from Google Search Volume Index data. In the post-2023 period, equal-weighted long-short portfolios yield monthly Fama-French five-factor alphas of 0.59-0.77%, with the premium concentrating among value and low-profitability firms. A decomposition shows that 54-66% of the spread reflects between-sector variation. Firms with higher AI exposure earn significantly lower market-adjusted abnormal returns on days of elevated public attention to AI regulation and safety. Taken together, these results suggest that AI risk is reflected in equity prices primarily through sector composition rather than through firm-level disclosure intensity, with within-sector pricing concentrated in selected industries.

Information

Lärosäte / institution
Handelshögskolan i Stockholm/Institutionen för finansiell ekonomi
Publiceringsdatum
2026
Uppsatstyp
Kandidat-uppsats
Språk
Engelska

Utforska vidare

Liknande uppsatser

Uppsatser med liknande ämnen och nyckelord.