“OpenAI proceeded to make a Scarlett Johansson-like voice for GPT-4o, even after she specifically told them not to, behaving in an entitled fashion that has angered many—highlighting their overall dismissive attitude towards artist consent.”
Our reading, the sentence that was checked:
OpenAI made a Scarlett Johansson-like voice for GPT-4o after she told them not to, showing dismissiveness toward artist consent.
Assessment: Mixed Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, What should we learn from OpenAI’s mistakes and broken promises?, paragraph 3.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-what-should-we-learn-from-openais-e20e39-a1
We recorded 15 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 2024-05-26. 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
- a Scarlett Johansson-like voice
- for GPT-4o
- even after she specifically told them not to
- angered many
- dismissive attitude towards artist consent
Our interpretation
- The factual core is recognizable resemblance plus creation of the voice after Johansson’s refusal.
- Making a voice is distinguished from subsequently demonstrating or releasing an existing voice.
- Declining to provide her voice is distinguished from expressly forbidding use of another actor’s similar voice.
- Entitlement and an overall attitude toward artists are judgments; they are not independently established by resemblance.
- The target period includes development, solicitation, refusal, and the May 2024 demonstration. The claim supplies no starting date.
Counts as supported if: Records establish recognizable resemblance and show that OpenAI created the relevant voice after Johansson expressly refused or objected to that use.
Counts as contradicted if: Reliable production records show that the voice was created before her refusal, or reliable evidence refutes the alleged resemblance. A mixed finding is appropriate if resemblance holds but the creation sequence fails.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Mixed
Part of the claim holds and part does not, and the record says which part is which.
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 resemblance component is supported, and Johansson declined participation before the GPT-4o demonstration. But the documented chronology places Sky’s original casting and recording before that refusal. Marcus’s wording blurs creating the voice with later using it. The records do not establish an express prior prohibition on a soundalike or prove the broader judgment about OpenAI’s attitude.
Limitations
- Resemblance is supported by reported perception, not a blinded acoustic study.
- The casting corroboration is available here through secondary reporting.
- Later work for GPT-4o is not fully documented in the public production account.
Evidence (4 items)
Supports the claim: NPR, carried by Iowa Public Radio,
like Johansson
Why it matters here: The report supports recognizable resemblance in the public demonstration. It does not establish cloning or deliberate imitation.
- Publisher
- NPR, carried by Iowa Public Radio
- Kind of source
- News report
- Describes events of
- Public from
Supports the claim: TechCrunch,
declined
Why it matters here: Confirms refusal of participation.
- Publisher
- TechCrunch
- Kind of source
- News report
- Describes events of
- Public from
Goes against the claim: OpenAI,
before any outreach
Why it matters here: OpenAI dates recording to June and July 2023 and Johansson’s first approach and refusal to September. This contradicts creation of the original Sky voice after her refusal, although deployment followed it.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Goes against the claim: Ars Technica,
after casting
Why it matters here: Reporting on documents, recordings, and the actor’s agent corroborates casting before outreach. This is secondary reporting of The Washington Post’s investigation, whose page could not be read directly.
- Publisher
- Ars Technica
- Kind of source
- News report
- Describes events of
- Public from
What would settle it
Dated production records and communications could establish whether later GPT-4o work deliberately introduced Johansson-like characteristics after a specific objection.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, What should we learn from OpenAI’s mistakes and broken promises?
OpenAI proceeded to make a Scarlett Johansson-like voice for GPT-4o, even after she specifically told them not to, behaving in an entitled fashion that has angered many—highlighting their overall dismissive attitude towards artist consent.
- Where
- paragraph 3
- Published
- 2024-05-26
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-what-should-we-learn-from-openais-e20e39-a1: Current. Outcome: Mixed. 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-what-should-we-learn-from-openais-e20e39-a1.