Of 307 documented AI incidents tied to 21 U.S.-listed companies in a peer-reviewed study, four were specifically identified in those companies' SEC filings within a year. The authors said investors relying on those reports could not reconstruct the documented incident record.

Samir Chincholikar and Robin Chawla, independent researchers in New York, used a pre-registered design for the study published October 21, 2026, in Frontiers in Artificial Intelligence, fixing and hashing the analysis plan before final estimation.

The Breakdown

AI-generated summary, reviewed by an editor. More on our AI guidelines.

Risk language

The Conference Board and ESGAUGE found that 72% of S&P 500 companies flagged AI as a material risk in annual filings available through August 15, 2025, up from 12% in 2023.

In the incident study, generic risk language covering the same technology or harm appeared for 278 events. A post-hoc analysis found only 84 matched both. Across the full sample, 70.4% had language matching only the technology, only the type of harm, or nothing at all.

The researchers started with the AI Incident Database's July 27, 2026 snapshot, selecting records dated from 2019 through 2026. They matched incidents to 21 U.S.-listed companies and searched Forms 8-K, 10-K and 10-Q filed in the year after each incident.

Beyond the four specifically identified incidents, three appeared through a related lawsuit or regulatory matter, bringing the total to seven. The remaining 300 had neither under the study's definitions.

Four is the pre-registered count. Re-adjudicating the three cases moved General Motors' OnStar Smart Driver case to specifically identified, giving five. A second coder disagreed on all three cases, so the authors treat seven as an upper bound, with substantive disclosure between five and seven of 307.

Alphabet accounted for 86 incidents, Meta for 81, Tesla for 36 and Microsoft for 35 in the study sample. Dropping the five companies with the most incidents left 53 incidents, and 92.5% of those still had neither form of disclosure.

Disclosed cases

Zillow and General Motors were among the specifically identified incidents. Zillow disclosed about $304 million in charges to wind down its algorithm-driven home-buying business. GM reported a $478 million Cruise restructuring adjustment tied to suspending its robotaxi operations. Substantive disclosures clustered where there were financial-statement effects or formal legal and regulatory consequences.

None of the sample's 70 misinformation and content incidents or 57 generative-AI error incidents received substantive disclosure. Other disclosures concerned litigation or regulatory proceedings, including Meta legal matters and Tesla NHTSA and Justice Department matters.

The materiality limit

The findings do not establish that any incident was legally material, required disclosure or involved noncompliance. Securities filings apply a materiality threshold.

Know someone who'd find this useful? ✉️ Email it to a friend in one click, or they can subscribe free here.

“Securities filings are not intended to be a complete repository of every operational incident related to a public company,” the authors wrote.

The database draws on media reports and misses internal or low-profile failures. The researchers did not search earnings calls, investor presentations, unfiled press releases or court dockets. An incident disclosed only through those channels counted as not specifically identified in this study.

Reporting rules

Cyber incidents carry a four-business-day clock under Form 8-K Item 1.05 once a company determines one is material. AI incidents had no equivalent duty during the study period. A study of cyber disclosures counted 161 incident filings, including amendments, from 117 companies between December 18, 2023 and October 21, 2026.

The EU AI Act's Article 73 requires providers of high-risk AI systems to report serious incidents to national market-surveillance authorities. Under the AI Omnibus, most high-risk obligations were delayed to December 2, 2027.

The authors propose re-running the same design under Article 73 or future U.S. guidance such as an Item 1.05-style rule. They call their figures a baseline for those comparisons.

Frequently Asked Questions

What did the study measure?

Samir Chincholikar and Robin Chawla took AI incidents from the AI Incident Database, matched 307 of them to 21 U.S.-listed companies, and searched those companies' Forms 8-K, 10-K and 10-Q for the 12 months after each incident.

How many incidents showed up in filings?

Four were specifically identified under the pre-registered coding, and five after the authors re-adjudicated the cases. Three more appeared only as a related lawsuit or regulatory matter, so substantive disclosure ranges from five to seven of 307.

Do companies talk about AI risk at all?

Yes. The Conference Board and ESGAUGE found 72% of S&P 500 companies flagged AI as a material risk in 10-Ks available through August 15, 2025. In the study, generic language on the same technology or harm appeared for 278 incidents, but only 84 matched both.

Does the study say companies broke disclosure rules?

No. The authors say the findings do not establish that any incident was legally material, required disclosure or involved noncompliance. The study also did not search earnings calls, investor presentations, unfiled press releases or court dockets.

Is there a reporting rule for AI incidents?

Not in the U.S. during the study period. Cyber incidents carry a four-business-day 8-K clock once a company deems one material. The EU AI Act's Article 73 requires serious-incident reports for high-risk systems, with most high-risk obligations delayed to December 2, 2027.

AI-generated summary, reviewed by an editor. More on our AI guidelines.

Anthropic pitches safety after system breaches; Microsoft opens the exit door
IMPLICATOR .ai Morning Briefing · From San Francisco   Tuesday, October 27, 2026 9 stops From San Francisco 1 The Editorial   Morning, humans. AI companies are putting
Anthropic Sells Its Fear of AI to Wall Street
A message appeared in a tip form for an unsolved homicide maintained by the Philadelphia Police Department. "I may have information regarding this case," it read. "I recall seeing someone matching the
OPINION: Why Anthropic's Warnings in Its IPO Filing Are Worthless
In 2019, Dario Amodei emailed housemates and close colleagues to say he had given up on making OpenAI work. Any progress the safety team made, he told them, was bound to be temporary and would always
AI Research AI News

San Francisco

Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: editor@implicator.ai