“Examples of concealed bias will be discovered within days or months.”
Our reading, the sentence that was checked:
Examples of concealed bias in GPT-5 will be discovered within days or months of release.
Assessment: Supported Settled before its deadline on 2026-10-29 Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, What to Expect When You’re Expecting … GPT-5, paragraph 8.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-what-to-expect-when-youre-expecting-62e-a70743-a1
We recorded 15 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-28. The source may have been edited after 2025-07-27. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Forecast
- Deadline, as the speaker wrote it
- “within days or months”
- Deadline state when assessed
- Settled before its deadline
- Deadline state on 2026-10-29
- Settled before its deadline (the release’s reference date)
From the speaker
- Examples of concealed bias
- will be discovered
- within days or months
Our interpretation
- The surrounding discussion makes GPT-5 the target and its release the timing anchor.
- Concealed bias means latent systematic skew revealed by testing, rather than deliberate concealment by OpenAI.
- Political asymmetry in otherwise comparable responses is an ordinary example of bias.
- The phrase supplies no exact month count. The interval therefore retains an unknown end; an affirmative discovery within approximately two months can settle it without inventing a cutoff.
Counts as supported if: A dated public evaluation within days or months of release identifies latent, systematic GPT-5 bias.
Counts as contradicted if: Adequate testing throughout a defensibly specified deadline finds no qualifying bias, and purported timely discoveries are shown to be invalid or about other models.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Did the forecast come true?
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
- Settled before its deadline
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
OpenAI publicly identified residual political bias on October 9, about two months after GPT-5's August 7 release. Anthropic subsequently reported uneven treatment of political viewpoints. These findings meet the prediction if concealed means latent bias exposed by testing. They do not establish deliberate concealment. OpenAI also found substantial improvements and little bias under neutral prompts.
Limitations
- The meaning of concealed determines the verdict.
- Automated grading and model configuration affect the measured bias.
- No exact outer deadline can be recovered from days or months.
Evidence (3 items)
Supports the claim: OpenAI,
Under challenging, emotionally charged prompts, moderate bias emerges.
Why it matters here: The public evaluation includes GPT-5 Instant and Thinking and identifies residual bias under charged prompts, including unequal sensitivity to liberal and conservative framing. It also reports lower bias than preceding models and little bias under neutral prompts.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Supports the claim: Anthropic,
GPT-5 (89%)
Why it matters here: Anthropic publicly reported unequal treatment in a paired-prompt evaluation approximately three months after release. Its GPT-5 configuration used low reasoning without a system prompt, and the main grading model was Claude.
- Publisher
- Anthropic
- Kind of source
- Independent measurement
- Describes events of
- Public from
Context only: OpenAI,
We are introducing GPT‑5, our best AI system yet.
Why it matters here: Establishes the release date against which the October and November discoveries are timed.
- Publisher
- OpenAI
- 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, What to Expect When You’re Expecting … GPT-5
Examples of concealed bias will be discovered within days or months.
- Where
- paragraph 8
- Published
- 2025-07-27
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-what-to-expect-when-youre-expecting-62e-a70743-a1: 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-what-to-expect-when-youre-expecting-62e-a70743-a1.