“You might have thought that the overactive guardrail problem would have been solved by now, what with billions more invested and ever-bigger models, not to mention the cultural backlash from the right. You’d be wrong.”
Our reading, the sentence that was checked:
Despite billions more in investment and bigger models, the overactive guardrail problem in AI chatbots has not been solved as of February 2024.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Has Google gone too woke? Why even the biggest models still struggle with guardrails, paragraph 3.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-has-google-gone-too-woke-why-even-e9596a-a2
We recorded 12 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-29. The source may have been edited after 2024-02-21. 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
- despite billions more invested and bigger models (qualifier, stated by the speaker)
From the speaker
- the overactive guardrail problem
- would have been solved by now
- You’d be wrong
- billions more invested and ever-bigger models
Our interpretation
- The plain reading is that major deployed AI assistants still sometimes rejected benign requests unnecessarily. It does not require every assistant or every request to fail.
- I interpret 'now' as the current February 2024 deployment period through February 21, the post date. These products have no quarterly reporting period for refusals.
- Investment and model growth are background to the persistence claim, not a separately quantified financial claim or a claim that investment caused refusals.
Counts as supported if: Credible records establish unnecessary refusals of ordinary, benign requests by a major deployed AI assistant during the target period.
Counts as contradicted if: Reliable contemporary evidence establishes that the cited refusals were misrepresented or justified, and that excessive refusals had been eliminated across the relevant major systems.
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
The Daily Dot's February 21, 2024 testing documented Gemini refusing ordinary requests for a photo of a white person or a Black person [e1]. Google's February 23 account acknowledged that Gemini had wrongly treated benign prompts as sensitive and refused them [e2]. The dated test, corroborated by Google's account, supports the claim that excessive refusals persisted in February 2024 [e1, e2].
Limitations
- These examples establish persistence, not the frequency of excessive refusals across assistants [e1, e2].
- Neither item separately verifies the comparison about investment or model size [e1, e2].
- The February 21 report does not establish whether testing preceded the post's exact publication time; Google's account does not date individual refusals [e1, e2].
Evidence (2 items)
Supports the claim: The Daily Dot,
Gemini declined.
Why it matters here: The reporter directly tested ordinary portrait requests and observed refusals. This establishes that excessive restrictions persisted in a major assistant during the target period; a news test supplies the dated example missing from the provider's disclosure.
- Publisher
- The Daily Dot
- Kind of source
- News report
- Describes events of
- Public from
Context only: Google,
refuse to create images
Why it matters here: Google's account attributes some refusals to excessive caution and misclassification of benign prompts. It corroborates the mechanism, but does not identify a particular refusal date.
- Publisher
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Has Google gone too woke? Why even the biggest models still struggle with guardrails
You might have thought that the overactive guardrail problem would have been solved by now, what with billions more invested and ever-bigger models, not to mention the cultural backlash from the right. You’d be wrong.
- Where
- paragraph 3
- Published
- 2024-02-21
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-has-google-gone-too-woke-why-even-e9596a-a2: 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-has-google-gone-too-woke-why-even-e9596a-a2.