“Lots of internal warnings were ignored.”
Our reading, the sentence that was checked:
OpenAI ignored many internal warnings about ChatGPT's effects on users' mental health.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, ChatGPT and delusions: an important new inside look at OpenAI, paragraph 4.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-chatgpt-and-delusions-an-important-7b33e1-a2
We recorded 2 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 2025-11-23. 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
- Lots of internal warnings
- were ignored
- Context: ChatGPT users experiencing mental health crises and reporting about OpenAI.
Our interpretation
- The plain reading is that OpenAI disregarded a substantial body of internal warnings about risks to ChatGPT users' mental health.
- Lots means many warnings, not merely more than one. Several testers discussing one launch can contribute evidence, but do not automatically establish this quantity. No numerical cutoff is supplied or imposed.
- Ignored means relevant decision-makers dismissed warnings or proceeded without meaningful responsive action at the relevant time. Later remedies do not erase earlier disregard; an imperfect response does not by itself establish disregard.
- The statement describes past conduct preceding the post, rather than a reporting quarter. The target period ends on the post date; its start remains unspecified because the supplied context gives no starting date.
- Warnings about other products provide context unless evidence connects them to decisions about ChatGPT.
Counts as supported if: Credible records or corroborated reporting establish a substantial number or sustained pattern of distinct internal mental-health warnings concerning ChatGPT, and show that OpenAI disregarded those warnings at the relevant decisions.
Counts as contradicted if: A sufficiently comprehensive record shows that the alleged body of ignored warnings was not substantial, did not concern ChatGPT's mental-health risks, or received meaningful responses at the relevant decisions. Examples of warnings receiving responses are insufficient by themselves to disprove that many others were ignored.
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
There is credible evidence that OpenAI set aside some internal concerns. The investigation describes the April launch and a delayed safety suggestion, while OpenAI acknowledges giving testers' concerns too little weight. But these records do not clearly establish the substantial body of disregarded warnings implied by “lots.” Several witnesses to one decision are not necessarily several distinct warnings. Subsequent safety work also does not disprove earlier disregard. The available evidence supports a narrower statement without settling Marcus's full claim.
Limitations
- The investigation was read through its full syndicated version because the original page was inaccessible.
- Reporting includes employee testimony whose underlying internal communications are not publicly reproduced.
- The quantity and meaning of ignored remain partly interpretive; insufficient evidence for lots is not evidence that few warnings existed.
- The historical conduct had occurred by the evaluation date, but the full claimed pattern was not established by the retrieved public record.
Evidence (3 items)
Supports the claim: The New York Times, syndicated by The Indian Express,
But when decision time came, performance metrics won out over vibes.
Why it matters here: The investigation reports that Model Behavior raised concerns about excessive validation before the release; elsewhere it describes a delayed break-reminder proposal and an employee's earlier concerns. This supports some disregarded warnings, but does not establish how many distinct warnings were ignored; underlying internal records are not reproduced.
- Publisher
- The New York Times, syndicated by The Indian Express
- Kind of source
- News report
- Describes events of
- Public from
Supports the claim: OpenAI,
we should’ve paid closer attention.
Why it matters here: OpenAI acknowledges launching despite expert testers' concerns, corroborating the April episode. It also says testers had not explicitly flagged sycophancy and describes a subsequent rollback, limiting what this episode establishes about explicit mental-health warnings and prolonged inaction.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Context only: OpenAI,
responds more appropriately and consistently
Why it matters here: OpenAI reports clinician-informed improvements addressing mental-health risks before Marcus's post. This weighs against interpreting the claim as complete or permanent inaction, but does not show that earlier warnings received timely responses.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
What would settle it
Dated internal communications or corroborated reporting establishing a substantial body of distinct ChatGPT mental-health warnings, who received them, and how OpenAI handled them. A comprehensive record showing timely responses to the alleged warnings could instead contradict the claim.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, ChatGPT and delusions: an important new inside look at OpenAI
Lots of internal warnings were ignored.
- Where
- paragraph 4
- Published
- 2025-11-23
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-chatgpt-and-delusions-an-important-7b33e1-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.
About assessment marcus-chatgpt-and-delusions-an-important-7b33e1-a2.
HeldBy this site
The openai review found the criteria unfair to the speaker's words.
About assessment marcus-chatgpt-and-delusions-an-important-7b33e1-a1, which was not published.