A study published on October 21 by Samir Chincholikar and Robin Chawla traces only seven of 307 examined AI cases into related US securities filings. Those seven include both descriptions of incidents and disclosures of directly related legal consequences.
The researchers matched AI Incident Database cases to 21 US-listed issuers and examined Forms 8-K, 10-K and 10-Q. Their source snapshot dates to July 27, 2026. Generic AI-risk language was much more common than information identifying the event.
The study measures visibility in specified filings, not all public communication. Company websites, press releases and other channels fall outside its full coverage. An absent match establishes neither unlawful nondisclosure nor deliberate concealment.
Selection from a media-derived database and issuer matching also limit the findings. Coders disagreed over legal-consequence classifications. Results cannot be generalized to all AI incidents; the broad AIID sample is not restricted to unauthorized agent actions.