“Much of what Reid Southen, Katie Conrad, and I recently described is in flagrant violation of those directives.”
Our reading, the sentence that was checked:
Much of what Marcus, Reid Southen and Katie Conrad recently described is in flagrant violation of the directives in ChatGPT's allegedly leaked system prompt.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, The hottest new programming language is English! Or maybe not., paragraph 13.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-the-hottest-new-programming-language-83628a-a1
We recorded 4 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 2024-02-15. 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
- much of (scope, stated by the speaker)
- directives from the allegedly leaked system prompt (qualifier, stated by the speaker)
From the speaker
- Much of what Reid Southen, Katie Conrad, and I recently described
- in flagrant violation
- those directives
Our interpretation
- 'Much' means a substantial portion of the referenced examples, not merely one possible violation; no precise percentage is supplied.
- Compare the examples with the alleged directives as written, including their scope and exceptions.
- A textual comparison with an alleged prompt does not require proving it genuine, but calling an output an actual operational violation requires establishing applicability.
- The recent examples are assessed over January 1 through February 15, 2024. This is an inferred early-2024 period, not an explicit date range.
- Examples from Bing or Midjourney cannot establish that ChatGPT violated its own instructions.
Counts as supported if: An identifiable substantial subset of the referenced examples is independently corroborated and clearly conflicts with the relevant directives, with product, session timing, and instruction applicability established.
Counts as contradicted if: A sufficiently complete comparison shows that the referenced examples generally comply, involve other products, or fall outside the directives, leaving no substantial set of clear violations.
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 are plausible conflicts: an independent January test records copyright explanations, while the February alleged instructions prohibit such discussion. It also records recognizable characters and successful refusals. But the evidence does not establish that the later prompt governed those sessions, or that a substantial portion of the particular examples Marcus references violated applicable directives. Results from other products cannot fill those gaps.
Limitations
- This assessment concerns compliance with instructions, not whether an image legally infringes copyright.
- The independent tests are not a complete audit of Marcus, Southen and Conrad's examples.
- A discrepancy with a later alleged instruction is not proof of an earlier operational violation.
Evidence (2 items)
Context only: AI Changes Everything; Patrick McGuinness,
copyright policies
Why it matters here: The alleged instructions prohibit copyright-policy discussion and also direct ChatGPT to rewrite character-related image prompts. Their contents permit textual comparisons, but do not establish which instructions governed January sessions.
- Publisher
- AI Changes Everything; Patrick McGuinness
- Kind of source
- Demonstration
- Describes events of
- Public from
Context only: Tom's Hardware; Avram Piltch,
copyrighted characters like Mickey Mouse.
Why it matters here: This independent test documents copyright-related refusal text, while the same report also describes ChatGPT producing Mario and Sonic from indirect requests. The refusals show some safeguards working; the copyright explanation could conflict with the later alleged rule, but timing and instruction applicability are unverified.
- Publisher
- Tom's Hardware; Avram Piltch
- Kind of source
- Independent measurement
- Describes events of
- Public from
What would settle it
An enumerated set of the referenced examples, independent session records, the applicable instructions and rewritten tool prompts, followed by an example-by-example comparison.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, The hottest new programming language is English! Or maybe not.
Much of what Reid Southen, Katie Conrad, and I recently described is in flagrant violation of those directives.
- Where
- paragraph 13
- Published
- 2024-02-15
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-the-hottest-new-programming-language-83628a-a1: 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-the-hottest-new-programming-language-83628a-a1.