149 public companies added AI risk disclosures this filing season.
RiskDiff extracts and diffs Item 1A risk factors from SEC 10-K filings — showing exactly what changed, word for word, across 431+ companies. Risk matching uses deterministic machine learning. AI interpretation is clearly labeled wherever it appears.
AI risk language appeared as a new addition in 149 out of 431 companies tracked — making it the second most common newly-added theme across recent 10-K filings. Companies include Alphabet, JPMorgan Chase, Meta, Microsoft, and Tesla, among many others.
GOOGLJPMMETAMSFTTSLAACNADBEAIG
Top catalysts this cycle — largest net risk shift ⚠ AI-graded
Ranked by net severity — sum of per-change scores (deterministic baseline + AI business-impact bump). Direction and themes come from the AI risk brief. The deterministic diff is on each company's page.
How it works — company-level filing diff
New disclosures
Risk sections that appear for the first time — what management now considers material enough to warn investors about.
Removed language
Text present in the prior filing but absent in the current one. Can indicate resolved exposure or a shift in framing.
Significant rewrites
Sections materially changed from year to year — matched by text similarity scoring across consecutive filings.
NVDA NVIDIA Corporation — 2024 vs 2023
+ U.S. export control restrictions have limited our ability to sell certain products in China and other markets, and further restrictions may materially affect our revenue.
MSFT Microsoft Corporation — 2024 vs 2023
+ AI-generated content could expose us to claims of defamation, intellectual property infringement, or other legal liability.
− Pandemics, epidemics, or outbreaks of disease may adversely affect our business, operations, and financial condition.
For AI agents & developers — MCP access
Ask your AI agent: “Which industrial companies newly disclosed tariff exposure in their most recent 10-K?”
RiskDiff exposes a live MCP server at https://riskdiff.com/api/mcp. Connect any MCP-compatible agent — Claude, Cursor, or your own pipeline — and query risk signals, company diffs, and filing changes directly. No scraping, no hallucination, data traced to source.