Passages Flagged as Likely Heavily AI-Written: S&P 500 10-K Risk Factors

## What this measures

The percentage of passages from Item 1A (Risk Factors) of S&P 500 annual reports that an open AI-text detector scores above a fixed cutoff, over each company's latest 10-K filed in the trailing 365 days.

## Why it matters

Language models can now write a whole passage of a script or a filing. If companies started handing that writing over wholesale, the share of flagged passages would climb well above the level human writing alone produces. This series is a tripwire for that shift. A separate Palanor measure, the words language models overuse (discourse.ai_vocabulary.* and filings.ai_vocabulary.*), tracks a lighter change: vocabulary picked up from edited or assisted text.

## How to read it

The cutoff was set so that 5% of a sample of passages from before ChatGPT (released 30 November 2022) cross it. Human writing alone therefore reads near 5% (across every 10-K in the baseline period the reading was about 5%). Read the series against that level, not against zero: a reading near it means no detectable change. A sustained move well above it means more passages read like model output; on its own it does not show how they were written (see Findings). Each value carries a 95% interval in its metadata.

## How it is made

10-Ks: S&P 500 members on the reading date, each one's latest 10-K filed in the trailing 365 days, Item 1A headings and text as parsed by Palanor. Risk-factor sections are largely carried from year to year, so the series moves slowly. Each document is cut into 300-word passages and up to four are picked by a fixed rule (a seeded shuffle keyed on the document), so the same document always gives the same passages. Each passage is scored by an open AI-text detector, TMR (Oxidane/tmr-ai-text-detector on Hugging Face, a RoBERTa classifier, MIT licence), at a pinned model revision (0ceddea903015ef99cbaa040a4d8a216aed9c683). The score is the model's probability that the passage is AI-written, reading the first 512 tokens. A passage is flagged when its score is above the cutoff: 0.9823, the 95th percentile of scores on risk-factor sections of S&P 500 10-Ks filed in 2022. Every passage counts once.

The detector was chosen in a test on 600 passages from these same calls and 10-Ks, each paired with a rewrite of the whole passage by a current Claude model. TMR told originals from rewrites with an AUROC of 0.91 on calls and 0.89 on 10-Ks. At a cutoff that flags 5% of the human passages, it caught 71% of the call rewrites and 66% of the 10-K rewrites. Another detector tested (an ensemble) failed on 10-K text and is not used. Changing the model revision requires re-running that test and the baseline; the cutoffs only hold for the revision they were fitted on.

## What it does not say

It counts passages flagged as likely heavily AI-written. It does not see light AI editing: a person's text that a model polished, or a model draft a person reworked, mostly scores as human. A flag on a single passage is a probability, not proof, and about one human passage in twenty is flagged anyway. That is why Palanor publishes only the share across hundreds of companies. It never shows a score for any company or document, and nothing here says that any company used AI.

## Findings — 2026-10-17

The first full scoring (2026-10-17) covered 2,421 S&P 500 10-Ks filed from July 2021 to October 2026. The share of flagged passages was about 5% in 10-Ks filed in 2022 and between 5% and 6% in every filing year since, with intervals that overlap the 2022 level: no clear change. Risk-factor sections are largely carried forward from one year to the next, so a change would show slowly. The earnings-call series, by contrast, rose (see discourse.ai_text.flagged_share).

## Source

Palanor calculation from Item 1A (Risk Factors) of annual reports on Form 10-K filed with the SEC (public domain). A value is published only when at least 150 companies have a scored 10-K in the window.