AI-generated misinformation became increasingly widespread, as Marcus warned in December 2022, but its effects have so far been less than he feared.
Mostly true True in substance, wrong in a detail.
Factual: Statements about the present or the past, checked against the record: did they hold up?
The second judge would not resolve this claim; the verdict shown is the first judge's.
What would count as true or false
There are three parts. (a) Marcus warned in December 2022 about AI-generated misinformation. (b) AI-generated misinformation then became increasingly widespread. (c) Its effects so far were smaller than he feared. Judge (c) against the harms he actually predicted in December 2022, such as large-scale disruption of trust or elections, not against any other bar. All three are judged as of when the retrospective statement was made. TRUE: a December 2022 warning by Marcus exists, credible reports or studies show AI-generated misinformation volume grew after it, and the documented impact up to the statement date fell short of the scale he warned of. MOSTLY TRUE: all three hold, but there is a small slip, such as the warning dating from late November or early January. MIXED: one of the three parts clearly fails while the others hold. Examples: no December 2022 warning can be found, or the documented impact roughly matched what he feared. MOSTLY FALSE: two parts fail. FALSE: the evidence shows AI-generated misinformation did not grow, and there was no such warning.
Set before the research began.
Evidence (4)
-
E1 AI's Jurassic Park moment Gary Marcus (Substack) · 10 December 2022 Own record
“Like it or not, these models are here to stay, and we as a society are almost certain to be overrun by a tidal wave of misinformation.”
-
E2 Experts: AI didn’t turn 2024 election, but threats persist PolitiFact · 19 December 2024
“But the anticipated avalanche of AI-driven misinformation never materialized.”
-
E3 Don’t Panic (Yet): Assessing the Evidence and Discourse Around Generative AI and Elections Knight First Amendment Institute at Columbia University · 7 July 2025
“There is little evidence to suggest that mainstream news organizations are using AI to create misleading content or fake news.”
-
E4 Tracking AI-enabled Misinformation: 3,749 AI Content Farm sites (and Counting), Plus the Top False Claims Generated by Artificial Intelligence Tools NewsGuard · 23 June 2026
“To date, NewsGuard’s team has identified 3,749 AI Content Farm news and information websites spanning 16 languages:”
Reasoning
E1 is Marcus's own 10 December 2022 post. In it he warned that society was "almost certain to be overrun by a tidal wave of misinformation," so part (a) holds. On part (b), NewsGuard counts 3,749 AI content-farm sites in 16 languages and describes the count as rising ("and Counting"), but the page gives no dated time series, so the evidence for growth is only partial (E4).
Cite this claim
Gary Marcus, “AI-generated misinformation became increasingly widespread, as Marcus warned in December 2022, but its effects have so far been less than he feared.” (said December 2022). Verdict: mostly true. Reliability API, claim c_0a53ce7e1848, method 0.1.0+c876030c, dataset release v1.1.2. https://epistacklabs.com/claims/c_0a53ce7e1848
This claim is in dataset release v1.1.2, with the same verdict.
Reply to this claim
More factual claims by Gary Marcus
-
Mostly true
“Even this was still a mix of euphemism (“attacks” were called “interactions”) and partial disclosure.”
OpenAI's latest account called attacks "interactions" and only partially disclosed what happened.
-
Mostly true
“It didn’t just attack a German web server.”
-
Resolved true
“It didn’t just attack Australia’s government servers.”
-
Mostly true
“So with LLMs, people do things like put stuff in a system prompt, and they can’t get systems with the system prompts to follow them.”
People put instructions in LLM system prompts but cannot get the systems to follow them.
-
Mostly true
“a just-published technique shows that current LLMs can be jailbroken quickly, and automatically.”
A newly published technique shows that current LLMs can be jailbroken quickly and automatically.
Verdicts are made automatically; how verdicts are made.