Pundit ClaimsAI claims by Gary Marcus and Ed Zitron, with the evidence

Gary Marcus Conditional forecast First said

“The more people use LLMs, the more trouble we are going to be in.”

When it comes to security, LLMs are like Swiss cheese — and that’s going to cause huge problems, paragraph 1. The full statement is at the source.

Our reading, the sentence that was checked:

The more people use LLMs, the more trouble we are going to be in.

Assessment: Unresolved No deadline stated on 2026-10-29 Evidence as of

Where and when Gary Marcus wrote it

Marcus on AI, When it comes to security, LLMs are like Swiss cheese — and that’s going to cause huge problems, paragraph 1.

Said
Source observed
live page, full text, public
Evidence as of
the cutoff for this assessment
Assessment recorded
assessment marcus-when-it-comes-to-security-llms-are-23d26c-a1

We recorded 7 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 2024-10-17. This page does not claim the wording was captured then.

What the claim commits to

Kind
Conditional forecast
Deadline, as the speaker wrote it
None stated. This site does not add one.
Deadline state when assessed
No deadline stated
Deadline state on 2026-10-29
No deadline stated (the release’s reference date)

Conditions

  • the more people use LLMs (antecedent, stated by the speaker)

From the speaker

  • The more people use LLMs
  • the more trouble we are going to be in

Our interpretation

  • In context, trouble means privacy breaches and security harms associated with LLM use.
  • The ordinary reading concerns increasing total harm, not necessarily increasing harm per user.
  • The antecedent requires evidence that LLM use expanded after the post. Generative-AI surveys are supporting evidence, but evidence identifying LLM applications is also needed.
  • No deadline, geographical scope, harm measure, or minimum increase is supplied.

Counts as supported if: After LLM use expands, reliable evidence establishes an accompanying increase in actual privacy or security harms attributable to that expansion.

Counts as contradicted if: Comparable evidence establishes that expanding LLM use did not increase the relevant aggregate harms over a sufficiently informative period. A short period without reported incidents would not suffice.

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

LLM use expanded, satisfying the condition. Netskope also reports increasing data-policy violations, which is consistent with the warning. But those measurements do not establish increased successful breaches or resulting harm caused by wider use. No decisive event settles the general relationship, and Marcus gave no deadline.

Limitations

  • Enterprise telemetry is not representative of all users.
  • The population survey measures generative AI more broadly than LLMs.
  • The meaning of trouble materially affects what would count as success.

Evidence (2 items)

  • Supports the claim: Federal Reserve Bank of St. Louis,

    work adoption increased from 33.3% to 37.4%.

    Why it matters here: The repeated survey documents increasing generative-AI adoption among working-age US adults. This supports the antecedent alongside the LLM-application measurements in e2, but does not measure security harm.

    Publisher
    Federal Reserve Bank of St. Louis
    Kind of source
    Independent measurement
    Describes events of
    Public from
  • Supports the claim: Netskope Threat Labs,

    The twofold increase in violations represents only the detected incidents.

    Why it matters here: Customer telemetry through October 2025 reports threefold user growth for applications including ChatGPT and Gemini, establishing expanded LLM use. Detected data-policy violations doubled, but these are not a verified count of successful breaches or resulting harm.

    Publisher
    Netskope Threat Labs
    Kind of source
    Independent measurement
    Describes events of
    Public from

What would settle it

A longitudinal investigation with comparable adoption and realized-harm measurements that establishes the claimed relationship, or persuasive evidence against it across substantial expansion.

Where it was said (1)

Said once in the sources this register covers.

  • : Marcus on AI, When it comes to security, LLMs are like Swiss cheese — and that’s going to cause huge problems

    The more people use LLMs, the more trouble we are going to be in.

    Published
    2024-10-17
    Source observed
    , live page
    Access
    full text, public

History

Assessment versions

Recorded events

  1. Assessment publishedBy this site

    First assessment published after both reviews and the final-text check passed.

    About assessment marcus-when-it-comes-to-security-llms-are-23d26c-a1.