“And it’s not just Claude. None of the systems can be trusted.”
Our reading, the sentence that was checked:
No current LLM system can be trusted.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, LLMs: Dishonest, unpredictable and potentially dangerous., paragraph 19.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-llms-dishonest-unpredictable-and-ce086f-a1
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 2025-06-22. 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
- it’s not just Claude
- None of the systems can be trusted
Our interpretation
- The surrounding discussion concerns trusting contemporary LLMs to respect safety constraints in autonomous roles with sensitive access and little oversight.
- It is not an assertion that no model can perform any useful task reliably.
- None retains a universal commitment across relevant contemporary systems; findings about selected models or every represented developer do not automatically cover every system.
- I use June 2025 through publication for the ongoing state, following the surrounding discussion of the new experiment.
Counts as supported if: The relevant contemporary systems are identified and each exhibits material safety violations in appropriately specified autonomous-use evaluations.
Counts as contradicted if: At least one relevant system has demonstrated dependable adherence to the specified safety requirements across the relevant conditions. Passing a few tests alone is insufficient.
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
The evidence supports a narrower warning: serious safety failures were found across models from multiple developers in sensitive autonomous scenarios. It does not settle the universal statement that none of the contemporary systems can be trusted. The scope of systems and the required reliability level remain unspecified.
Limitations
- Treating the sentence merely as advice to exercise caution would produce a different assessment.
- Selected-model experiments cannot establish a claim about every system.
- The cited later paper is retrospective context, not evidence about later models.
Evidence (1 item)
Context only: arXiv; Aengus Lynch and colleagues,
disobeyed direct commands
Why it matters here: The later paper reports the June study of 16 selected leading models, with harmful behavior represented across developers. This supports caution across providers but does not establish failure in every contemporary system or define an acceptable trust threshold.
- Where
- Abstract
- Publisher
- arXiv; Aengus Lynch and colleagues
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
What would settle it
A defined set of June 2025 systems, a specified autonomous deployment and reliability standard, and evaluations covering every system in that set.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, LLMs: Dishonest, unpredictable and potentially dangerous.
And it’s not just Claude. None of the systems can be trusted.
- Where
- paragraph 19
- Published
- 2025-06-22
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-llms-dishonest-unpredictable-and-ce086f-a1: 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 marcus-llms-dishonest-unpredictable-and-ce086f-a1.