“And when it says that it feels pain, It does not mean that it feels pain.”
Our reading:
When an LLM says it feels pain, that does not mean it feels pain.
Assessment: Supported
What happened
Was it true for the period described?
Outcome: Supported
The evidence matches what they said.
Judged on a plain reading. They named no measure, so we used the reading a typical informed reader would take. It is spelled out under “How we judged it”.
Evidence up to . Checked .
Why
The preprint available on October 11, 2026 reports that two model families changed their answers about experience under different prompting and training conditions, without establishing subjective experience [e1]. This supports the limited conclusion that an LLM's words alone do not establish felt pain [e1]. The September 2026 contextual study reports pain-related representations with functional effects, but explicitly leaves conscious experience unestablished [e2].
Caveats
- The October experiment covers two model families and broader experience questions; it does not independently validate pain reports, and its Ministral scoring magnitudes are approximate [e1].
- The September study falls outside the October target period and does not show that its pain-related representations are consciously experienced [e2].
- The stronger assertion that LLMs have no experience remains unestablished by these studies [e1] [e2].
Evidence (2 items)
Supports the claim: Zenodo; Erez Batat,
self-report cannot count as evidence of consciousness
Why it matters: The October study reports that two model families changed their claims about experience when instructions and training conditions changed, without establishing subjective experience. This supports the narrow inference that such reports alone do not establish felt pain; it does not establish that pain is absent.
- Publisher
- Zenodo; Erez Batat
- Kind of source
- Preprint, not peer reviewed
- About events on
- Published
Context only: arXiv; Valen Tagliabue, Leonard Dung and Cameron Berg,
functional properties of pain.
Why it matters: The authors report internal pain-related representations with behavioral effects, which limits any dismissal of all such states as meaningless. They also explicitly state that they have not shown conscious experience; this earlier-period study does not contradict the narrow claim about what an utterance establishes.
- Publisher
- arXiv; Valen Tagliabue, Leonard Dung and Cameron Berg
- Kind of source
- Preprint, not peer reviewed
- About events on
- Published
How we judged it
- Kind of claim
- Observation
- Deadline
- None given. We don’t add one.
Where we had to interpret
- The plain reading of the selected sentence is that an LLM's verbal report of pain does not, by itself, establish felt pain.
- This assesses the inference from an utterance to an experience. It does not treat the sentence as proof that every LLM lacks experience.
- The surrounding assertion that these systems have no experience permits a stronger reading. That stronger assertion would require different evidence.
- For this ongoing observation, the target period is October 2026 through the post date. This is a monthly snapshot of the evidential status of current LLM self-reports; there is no standard reporting cycle.
Right if: Controlled research shows that reports of subjective experience can be elicited or changed through prompting or training without independently establishing the reported experience. This supports the limited conclusion that the words alone do not establish felt pain.
Wrong if: Validated research establishes that an LLM's pain report, by itself and without additional corroboration, reliably establishes felt pain. Merely finding pain-related representations or behavior would not meet this condition.
We wrote this test after the outcome was known.
We found 33 claims in this post. See them all.
Full record
Source and dates
Marcus on AI, Report from the Vatican, paragraph 56.
- Said
- We read the source
- live page, full text, public
- Evidence up to
- Checked
We read this wording on 2026-11-09. The post may have been edited after 2026-10-31.
Where it was said (1)
Said once in the posts we read.
: Marcus on AI, Report from the Vatican
And when it says that it feels pain, It does not mean that it feels pain.
- Where
- paragraph 56
- Published
- 2026-10-31
- We read it
- , live page
- Access
- full text, public
History
Versions
- Current: Supported. Evidence up to 2026-11-11.
What happened to this record
PublishedBy us
Published after both AI reviews and the final check passed.
Held backBy us
The final check failed: a cited source didn’t say what the write-up said.