“Unless and until somebody can invent a new architecture that can reliably track provenance of generative text and/or generative images, infringement – often not at the request of the user — will continue.”
Our reading, the sentence that was checked:
Unless a new architecture that reliably tracks provenance is invented, generative AI infringement, often unrequested by users, will continue.
Assessment: Unresolved No deadline stated on 2026-10-29 Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Things are about to get a lot worse for Generative AI, paragraph 27.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-things-are-about-to-get-a-lot-worse-395723-a2
We recorded 14 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 2023-12-29. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Conditional forecast
- Deadline, as the speaker wrote it
- None stated. This site does not add one.
- Deadline state when assessed
- No deadline stated
- Deadline state on 2026-10-29
- No deadline stated (the release’s reference date)
Conditions
- unless a new architecture reliably tracking provenance is invented (antecedent, stated by the speaker)
From the speaker
- Unless and until
- invent a new architecture
- reliably track provenance
- generative text and/or generative images
- infringement
- often not at the request of the user
- will continue
Our interpretation
- The forecast begins on 2023-12-29 and has no stated deadline.
- Provenance means tracing generated material to underlying sources, rather than identifying the generating model.
- The antecedent is the continued absence of an invented architecture meeting that capability. The wording says invention, not merely commercial deployment.
- “Often” requires evidence of a recurring unintended-output problem; one deliberately elicited example cannot establish its frequency.
- Copyright infringement requires a legal basis. Recognizable character generation alone is evidence of risk, not an automatic legal finding.
Counts as supported if: While the absence of a qualifying provenance architecture is established, dated evidence shows continuing infringement and recurring outputs that users did not request to copy.
Counts as contradicted if: While that antecedent remains established, compelling evidence shows infringement has been eliminated across the relevant systems without such an architecture.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Did the forecast come true?
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 deadline stated
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
On November 11, 2025, the Munich court reported output-related copyright infringement in a specific case; its judgment was not final when announced [e2]. The study available on March 26, 2025 shows recognizable character generation without explicit character names, but does not establish recurring unintended legal infringement [e1]. The August 3, 2026 paper describes provenance limitations in code generation, but does not establish that no qualifying text or image architecture had been invented [e3]. These records do not establish the combination of infringement, recurring unintended copying and the provenance condition required for early resolution [e1, e2, e3].
Limitations
- Selected experimental prompts do not establish the frequency of unintended infringing outputs in ordinary use [e1].
- The court finding concerns a specific German case; subsequent appellate status is not established [e2].
- Code-generation research does not establish the absence of a qualifying provenance invention across text and image systems [e3].
Evidence (3 items)
Supports the claim: Landgericht München I, official court press release,
durch die Outputs
Why it matters here: The court reported that the output-related infringement was not covered by a statutory exception. This supports continued infringement in a specific German case, but does not establish unintended-use frequency or the provenance antecedent.
- Publisher
- Landgericht München I, official court press release
- Kind of source
- News report
- Describes events of
- Public from
Context only: arXiv; Luxi He and colleagues,
indirect anchoring
Why it matters here: The study documents a mechanism through which generic prompts produce recognizable characters. It supports the continued-risk component, but deliberately selected experimental prompts do not measure how often ordinary users unintentionally infringe.
- Publisher
- arXiv; Luxi He and colleagues
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: arXiv; Alejandro Velasco and colleagues,
Existing mitigations
Why it matters here: This research describes unresolved provenance problems in code-generation tools. It does not establish that no qualifying text or image architecture had been invented anywhere.
- Publisher
- arXiv; Alejandro Velasco and colleagues
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
What would settle it
Dated evidence establishing the absence of a qualifying provenance architecture, together with legally grounded infringement and recurring unintended copying during that period; alternatively, compelling evidence that infringement was eliminated across the relevant systems without such an architecture.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Things are about to get a lot worse for Generative AI
Unless and until somebody can invent a new architecture that can reliably track provenance of generative text and/or generative images, infringement – often not at the request of the user — will continue.
- Where
- paragraph 27
- Published
- 2023-12-29
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-things-are-about-to-get-a-lot-worse-395723-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-things-are-about-to-get-a-lot-worse-395723-a2.