Can Language Models Identify Shadow Trading Targets?

Summary

Shadow trading — trading in a peer firm's securities on material nonpublic information about an "economically linked" company — is a novel and contested theory of insider trading liability, first prosecuted in SEC v. Panuwat (2023). Enforcing it requires identifying economically linked firms ex ante, a determination the SEC currently makes only after the fact using mass market surveillance infrastructure. In this paper, we ask whether NLP can do what the theory presumes insiders already do: identify peer firms in advance from publicly mandated disclosures. Using a two-stage LLM pipeline over the Management's Discussion and Analysis sections of SEC 10-K filings, we score semantic similarity across 30 M&A events in five industries and relate it to announcement-day abnormal returns. On the Panuwat fact pattern itself the pipeline recovers Incyte among the closest peers, but across the full dataset we find no association: the within-event rank correlation is +0.07 (permutation p = 0.37), and the mean per-event Spearman correlation is +0.05 with a confidence interval narrow enough to exclude any moderate relationship rather than merely fail to detect one. Read case by case, 14 of 30 events support the hypothesis, 12 contradict it, and 4 are ambiguous. These results are exploratory and bound to this pipeline and corpus, but they put pressure on the empirical premise of shadow trading enforcement and bear on constitutional questions surrounding the SEC's financial surveillance infrastructure.

Recommended Citation

Wilson, S., MacKay, M., Marello, A., & Bhattacharyya, T. (2026). Can Language Models Identify Shadow Trading Targets? An NLP Evaluation of SEC Enforcement Theory. AI for Law Workshop, ICML 2026.

BibTeX

@inproceedings{wilson2026shadow,
  title={Can Language Models Identify Shadow Trading Targets? An NLP Evaluation of SEC Enforcement Theory},
  author={Wilson, Sarah and MacKay, Michael and Marello, Anthony and Bhattacharyya, Trinav},
  booktitle={AI for Law Workshop at ICML 2026},
  eprint={2608.01322},
  archivePrefix={arXiv},
  year={2026}
}