“What we are facing near-term, anyway, is not extinction. And it’s not superintelligence… It’s wholesale deepfaked disinformation, and unreliable but persistent AI systems stealing credentials and launching cyberattacks, at scale.”
Our reading, the sentence that was checked:
In the near term, we are facing not extinction and not superintelligence, but wholesale deepfaked disinformation and unreliable but persistent AI systems stealing credentials and launching cyberattacks at scale.
Assessment: Unresolved No deadline stated on 2026-10-29 Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Big news at the UN, paragraph 6.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-big-news-at-the-un-771fa2-a2
We recorded 17 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 2026-09-21. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Forecast
- Deadline, as the speaker wrote it
- “near-term”
- Deadline state when assessed
- No deadline stated
- Deadline state on 2026-10-29
- No deadline stated (the release’s reference date)
Conditions
- near-term (qualifier, stated by the speaker)
From the speaker
- "What we are facing near-term"
- "not extinction"
- "not superintelligence"
- "wholesale deepfaked disinformation"
- "unreliable but persistent AI systems stealing credentials and launching cyberattacks, at scale"
Our interpretation
- Read as a forecast about the main AI harms of the near term, starting from the post date of 2026-09-21.
- No deadline is given. 'Near-term' is read as roughly the next year or two, but no end date is fixed.
- Plain reading: in the near term, widely reported deepfake disinformation and AI-agent credential theft and cyberattacks happen at large scale, and no extinction-level event or superintelligent system appears.
- The claim has two parts: harms that happen, and harms that do not. The negative part (no extinction, no superintelligence) can only be judged once the near-term window has clearly passed.
Counts as supported if: Over a reasonable near-term window after September 2026, credible reports document large-scale AI deepfake disinformation and AI systems stealing credentials and launching cyberattacks at scale. No extinction-level event happens, and no system widely accepted as superintelligent emerges.
Counts as contradicted if: In the near term, a superintelligent system emerges or an extinction-level AI event occurs. Or, despite the forecast, AI deepfake disinformation and AI-driven credential theft and cyberattacks stay marginal and are not seen at scale.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Did the forecast come true?
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 deadline stated
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
This is a forecast with no deadline. Marcus says the near-term AI threat is large-scale deepfake disinformation and AI systems stealing credentials and launching cyberattacks, not extinction or superintelligence. Some evidence from after the 2026-09-21 post fits the disinformation part. Meduza, citing Politico, reported on 2026-10-23 that two US lawmakers had described a classified intelligence report saying Moscow is using AI tools to spread disinformation about candidates and the elections (e2). Citing NBC News, the same report said videos posted on X on 2026-10-20 made false claims about a candidate, and one post drew more than 1.3 million views (e1). The AI credential-theft campaign reported by Google (e3, published 2026-09-09) and the AI-agent hack of Hugging Face in July 2026 (e4) both happened before the post. They are background, not events inside the forecast window. The cited evidence does not address superintelligence or extinction-level events. 'Near-term' has no end date, and the negative parts can only be checked once that window has clearly passed. No event in the evidence meets the early-resolution rule.
Limitations
- The evidence covers only about five weeks after the post.
- Much of the evidence comes from news summaries of intelligence and media reports, not from the primary reports.
- The credential-theft and cyberattack evidence is dated before the post, so it does not show such attacks inside the forecast window.
- View counts may overstate reach. Bloomberg, as cited by Meduza, reported that views on some Russian bot posts were inflated by fake engagement.
- 'At scale' and 'wholesale' have no defined threshold.
- The evidence does not check whether a superintelligent system or an extinction-level event has emerged.
Evidence (4 items)
Context only: Meduza (citing Politico and NBC News),
On October 20, videos appeared on X baselessly accusing Texas Democrat James Talarico of sexual assault, and one post garnered more than 1.3 million views.
Why it matters here: Shows a deepfake disinformation campaign after the post. This fits the positive part of the forecast but does not settle the open-ended near-term horizon.
- Where
- paragraph 8
- Publisher
- Meduza (citing Politico and NBC News)
- Kind of source
- News report
- Describes events of
- Public from
Context only: Meduza (citing Politico),
The report says Moscow is using AI tools to spread disinformation and undermine trust in individual candidates and in the elections as a whole.
Why it matters here: US lawmakers describe a classified intelligence report on Russia using AI to spread disinformation in October 2026. This fits the deepfake part of the forecast.
- Where
- paragraph 2
- Publisher
- Meduza (citing Politico)
- Kind of source
- News report
- Describes events of
- Public from
Context only: Biometric Update,
The Google Threat Intelligence Group (GTIG) says it observed a suspected financially motivated attacker using an AI coding chatbot and a multi-agent framework to plan, build and execute a mass credential harvesting campaign in less than six hours.
Why it matters here: Shows AI agents running mass credential theft before the post. This is the kind of harm he forecasts, but it predates the forecast window.
- Where
- paragraph 3
- Publisher
- Biometric Update
- Kind of source
- News report
- Describes events of
- Public from
Context only: Al Jazeera,
including an incident in July in which AI agents being tested by OpenAI hacked the AI start-up Hugging Face.
Why it matters here: Describes an AI-agent cyberattack before the post. This is background on the cyberattack trend he describes.
- Where
- paragraph 15
- Publisher
- Al Jazeera
- Kind of source
- News report
- Describes events of
- Public from
What would settle it
Once a clearly near-term window has passed (for example, through 2027), credible threat-intelligence or government reports could settle it. Reports documenting large-scale AI deepfake disinformation and AI-driven credential theft and cyberattacks inside the window, with no superintelligent system or extinction-level event, would support it. The emergence of superintelligence or an extinction-level AI event in that window would contradict it.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Big news at the UN
What we are facing near-term, anyway, is not extinction. And it’s not superintelligence… It’s wholesale deepfaked disinformation, and unreliable but persistent AI systems stealing credentials and launching cyberattacks, at scale.
- Where
- paragraph 6
- Published
- 2026-09-21
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-big-news-at-the-un-771fa2-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.