Whose Risks Count? What 1,901 AI Incidents Reveal About Corporate Disclosures
November 20, 2026
An analysis of 1,901 AI incidents matched to SEC filings shows that companies focus disclosure on technical failures rather than social harms. Half of the analyzed incidents involved risk types that the companies had never previously disclosed in their financial filings.
HOW THIS AFFECTS YOU
●
founderYou might not be legally required to disclose social harms unless they become material financial risks.
●
investorYou can see that most social harms from AI are not appearing in corporate risk disclosures.
●
policyThis shows a gap between technical failure reporting and the reporting of social impacts.