Corporate AI Disclosure from US 10-K Filings, 2010-2021

Replication Data for 'How Does the Stock Market Interpret Non-tech Firms’ AI-related Disclosures?' contains corporate AI disclosures extracted from US 10-K filings between 2010 and 2021. The dataset was created by authors Enache, Luminita; Impink, Johannes; Lyubimov, Alexey; and Moldovan, Rucsandra, and is associated with a paper under review as of 2026. It is hosted on the Borealis Harvested Dataverse platform.

Use Cases

• Analyze the relationship between AI disclosure and stock market reaction based on the described financial text data.
• Study the evolution of AI-related language in corporate annual reports over the 2010-2021 time period.
• Train NLP models to identify and classify AI-related disclosures in financial documents based on the described text corpus.
• Conduct event studies to measure investor response to AI announcements based on the described disclosure data.

Strengths

• Data covers a 12-year period from 2010 to 2021, providing longitudinal analysis potential.
• Source documents are official US 10-K filings, a standardized and authoritative corporate disclosure format.
• Dataset is explicitly linked to a specific, citable research paper (SSRN abstract 6374659).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for large-scale modeling.
• Description metadata is limited; actual data quality and completeness require manual inspection after download.

Provenance

Source

US Securities and Exchange Commission (SEC) 10-K filings.

Collection Method

Likely extracted via text analysis of corporate annual reports.

Time Range

2010 to 2021

Freshness

Last updated 2026-07-18 04:10:37; freshness should be verified.

Geography

United States