“As I wrote (and broadcast) recently, LLMs are running out of training data, and the desperation to find more will lead to significant legal consequences when it's revealed exactly how much they stole and from whom they stole it from.”
Our reading, the sentence that was checked:
LLMs are running out of training data.
Assessment: Unresolved Evidence as of
Where and when Ed Zitron wrote it
Where's Your Ed At, They're Looting The Internet, paragraph 52.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment wyea-the-great-looting-of-the-internet-7b3c4b-a2
We recorded 21 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-28. The source may have been edited after 2024-04-15. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Observation
- Deadline, as the speaker wrote it
- None stated. This site does not add one.
- Deadline state when assessed
- No horizon
- Deadline state on 2026-10-29
- No horizon (the release’s reference date)
From the speaker
- LLMs are running out of training data
Our interpretation
- This assessment covers the supplied normalized proposition about data availability. It does not grade the separate forecast of legal consequences in the same sentence.
- The ordinary reading is that frontier LLM training was approaching a shortage of useful additional human-generated text at prevailing scaling rates, rather than that every possible training source was already exhausted.
- A projection can inform that assessment, but its assumptions and distinction between high-quality text and all usable data must be preserved.
- No period is stated. I use 2024 through publication as the current-state reporting window.
Counts as supported if: Stock-and-demand estimates applicable to the target period show useful additional training text approaching exhaustion at prevailing frontier training rates.
Counts as contradicted if: Comparable estimates applicable to the target period show that usable new data was not approaching a constraint under then-current training practices.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Unresolved
The record was checked, and the evidence available by the cutoff does not settle it.
Scored on a plain reading. The speaker named no measure for this claim, so it was judged on the reading an informed reader would take, stated below under our interpretation. Both reviews had to find that reading fair.
- Evidence as of
- Deadline state when assessed
- No horizon
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
Epoch's November 10, 2022 analysis projects data constraints if trends continue, but describes its uncertainty as very high and identifies data efficiency and synthetic data as factors its model should incorporate [e1]. Its June 6, 2024 update places full utilization of usable human-generated public text between 2026 and 2032 with 80% confidence, depending on its assumptions about data and training [e2]. The update also revises the earlier methodology [e2]. Neither projection settles whether useful additional text was approaching exhaustion during January 1–April 15, 2024 [e1, e2].
Limitations
- The earlier analysis supplies a conditional projection, not a measurement of the target-period condition [e1].
- The June update postdates the target period and changes the assumptions underlying the earlier estimate [e2].
- The public-text estimate does not establish a shortage across all possible training sources [e2].
Evidence (2 items)
Context only: Epoch AI,
the model uncertainty is very high.
Why it matters here: The study projected high-quality language-data exhaustion before 2026, while projecting much later limits for lower-quality language data. It explicitly qualified those projections and did not measure an April 2024 shortage.
- Where
- Results
- Publisher
- Epoch AI
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: Epoch AI,
2032.
Why it matters here: The later update projected a future constraint and revised the earlier high-quality-data estimate. Its changed methodology cautions against treating the earlier projection as an established current shortage.
- Publisher
- Epoch AI
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
What would settle it
A stock-and-demand comparison applicable to January 1–April 15, 2024 showing whether useful additional human-generated text was approaching exhaustion at prevailing frontier training rates, with consistent assumptions about quality, language, deduplication, repetition and access.
Where it was said (1)
Said once in the sources this register covers.
: Where's Your Ed At, They're Looting The Internet
As I wrote (and broadcast) recently, LLMs are running out of training data, and the desperation to find more will lead to significant legal consequences when it's revealed exactly how much they stole and from whom they stole it from.
- Where
- paragraph 52
- Published
- 2024-04-15
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment wyea-the-great-looting-of-the-internet-7b3c4b-a2: Current. Outcome: Unresolved. Evidence as of 2026-10-28. Public date not recorded.
Recorded events
Assessment publishedBy this site
First assessment published after both reviews and the final-text check passed.
About assessment wyea-the-great-looting-of-the-internet-7b3c4b-a2.
HeldBy this site
The final-text check failed: citations_correct.
About assessment wyea-the-great-looting-of-the-internet-7b3c4b-a1, which was not published.