A Frontiers in Artificial Intelligence study by S. Chincholikar and R. Chawla found that only four out of 307 corporate AI incidents were specifically identifiable in subsequent SEC filings.
The study by S. Chincholikar and R. Chawla found that seven out of 307 matched corporate AI incidents, or 2.3%, had substantive disclosures detailing legal or regulatory consequences.
Researchers S. Chincholikar and R. Chawla determined that 300 out of 307 analyzed AI incidents formed an "AI incident disclosure shadow," accounting for 97.7% of the sample.
A study by S. Chincholikar and R. Chawla found generic risk language in SEC filings accompanying 278 out of 307 analyzed AI incidents, representing 90.6% of the sample.
In SEC filings analyzed by S. Chincholikar and R. Chawla, Zillow reported approximately $304 million in charges to wind down its algorithm-driven home-buying business.
In SEC filings analyzed by S. Chincholikar and R. Chawla, General Motors reported a $478 million Cruise restructuring adjustment connected to autonomous-vehicle operations.
Analysis by S. Chincholikar and R. Chawla revealed that 216 out of 307 AI incidents, or 70.4%, were accompanied in SEC filings by at most weakly related language.
The study by S. Chincholikar and R. Chawla found zero substantive SEC disclosures across 70 misinformation or content-related incidents and 57 generative-AI error incidents.
Reviewers in the study by S. Chincholikar and R. Chawla examined the Risk Factors and Legal Proceedings sections of the first annual report following each AI incident.
The study by S. Chincholikar and R. Chawla evaluated a July 27, 2026 snapshot of 1,389 records dated from 2019 through 2026 in the AI Incident Database.
Researchers S. Chincholikar and R. Chawla narrowed their analytical sample to 307 AI incidents associated with 21 U.S.-listed domestic corporate issuers.
Across multiple robustness specifications evaluated by S. Chincholikar and R. Chawla, the estimated AI incident disclosure shadow rate consistently ranged between 92.5% and 99.0%.
Alphabet with 86 incidents, Meta with 81, Tesla with 36, and Microsoft with 35 comprised the main concentration of the analytical sample in the study by S. Chincholikar and R. Chawla.
Under the framework defined by S. Chincholikar and R. Chawla, Tier 1 required incident-specific disclosure that provided sufficient factual detail to link the filing to the AI Incident Database record.
Under the framework defined by S. Chincholikar and R. Chawla, Tier 2 covered formal legal or regulatory consequences linked to an incident without describing the underlying AI event.
Under the framework defined by S. Chincholikar and R. Chawla, Tier 3 comprised generic risk language naming technology or harm families, while Tier 4 indicated no related filing language was located.
Researchers S. Chincholikar and R. Chawla searched SEC Forms 8-K, 10-K, and 10-Q filed by issuers during the 12 months following each recorded AI incident date.
In an exploratory severity analysis, S. Chincholikar and R. Chawla observed substantive SEC disclosure in two of 42 severe cases, one of 50 moderate cases, and four of 215 limited cases.
Removing the five largest incident contributors from the sample yielded a 92.5% AI incident disclosure shadow estimate in the study by S. Chincholikar and R. Chawla.
A pre-registered issuer-clustered bootstrap calculation in the study by S. Chincholikar and R. Chawla produced an AI incident disclosure shadow rate of 97.5%.
The study by S. Chincholikar and R. Chawla did not analyze non-SEC filing communications such as earnings calls, investor presentations, corporate websites, or press releases.
The research sample of S. Chincholikar and R. Chawla excluded 1,082 database records involving private companies, foreign-listed issuers, delisted issuers, or unresolved organizations.
Authors S. Chincholikar and R. Chawla explicitly stated that their study makes no legal determination regarding materiality, formal disclosure duties, or legal compliance.