“Gary has written here endlessly about the troubles with reliability, apparently inherent, in LLMs. If you write code with an LLM, you are asking for trouble; the kind of garden-variety hallucinations that Gary has described in, for example, biographies, have parallels in LLM-generated code.”
Our reading, the sentence that was checked:
LLMs have apparently inherent reliability problems, and code written with them contains hallucinations, making LLM coding risky.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, LLMs + Coding Agents = Security Nightmare, paragraph 4.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-llms-coding-agents-security-nightmare-147dcd-a3
We recorded 12 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 2025-08-17. 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)
Conditions
- when writing code with an LLM (qualifier, stated by the speaker)
From the speaker
- troubles with reliability, apparently inherent, in LLMs
- If you write code with an LLM, you are asking for trouble
- hallucinations ... have parallels in LLM-generated code
Our interpretation
- The plain reading is that LLM-generated code has recurring fabricated dependencies or other invented references that create practical reliability risks.
- Apparently inherent expresses a tentative explanation. It is not treated as a firm claim that every possible future system must fail.
- The statement describes technology available at publication. With no standard reporting quarter for this subject, the observation window is the current calendar year through 2025-08-17.
- Asking for trouble means exposure to a real risk, not that every generated program fails.
Counts as supported if: Systematic evaluations in the observation period find recurring hallucinated code dependencies or references, with credible consequences for execution or security.
Counts as contradicted if: Representative evaluations find that the claimed hallucinations are absent or practically negligible across ordinary code-generation tasks.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Supported
The dated evidence matches what the claim committed to, under the criteria frozen before grading.
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
Trend Micro's report, available June 5, 2025, found that foundation models mostly produced no fabricated package names across 100 web-development tasks, but invented names on some complex prompts [e1]. Coding agents were tested on the ten tasks with the highest foundation-model hallucination rates; reasoning and live search approximately halved hallucinations, while some invented names persisted [e1]. Live validation further reduced hallucinations without eliminating them in the tested tasks [e1]. These findings support a practical reliability risk from fabricated dependencies and show that safeguards can reduce that risk [e1].
Limitations
- The report was available within the January 1–August 17, 2025 observation window, but exact experiment dates are not provided [e1].
- The selected agent tasks do not establish average developer risk or representative production failure rates [e1].
- The findings do not establish that hallucinations are architecturally inevitable, that every generated program fails, or that malware infections occurred [e1].
Evidence (1 item)
Supports the claim: Trend Micro,
occasional spikes persist
Why it matters here: The reported comparison found invented dependencies in foundation models and coding agents, supporting a concrete reliability and supply-chain risk. It also found predominantly clean foundation-model outputs and substantial reductions with agent reasoning and validation, which limit any claim of pervasive or unavoidable failure.
- Publisher
- Trend Micro
- Kind of source
- Independent measurement
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, LLMs + Coding Agents = Security Nightmare
Gary has written here endlessly about the troubles with reliability, apparently inherent, in LLMs. If you write code with an LLM, you are asking for trouble; the kind of garden-variety hallucinations that Gary has described in, for example, biographies, have parallels in LLM-generated code.
- Where
- paragraph 4
- Published
- 2025-08-17
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-llms-coding-agents-security-nightmare-147dcd-a3: Current. Outcome: Supported. 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-coding-agents-security-nightmare-147dcd-a3.
HeldBy this site
The final-text check failed: citations_correct.
About assessment marcus-llms-coding-agents-security-nightmare-147dcd-a2, which was not published.