Artificial Intelligence in fraud detection: textual analysis of 10-K filings

In this paper, we investigate the potential of Artificial Intelligence (AI) in detecting fraud by analyzing linguistic indicators in 10-K filings. We analyze the word frequencies (positive, negative, uncertainty, litigious), consistency, and readability in the MD&A sections. The AI model, BERT, was trained on these factors to predict fraud, showing significant promise compared to traditional models. The findings suggest that fraudulent filings tend to have more positive words, inconsistent language, and higher readability. This highlights AI’s practical role in improving fraud detection in financial reports.